← RevenueBench home
RevenueBench · Model Reference Guide · Rev 2026-09-28
Every Model We Tested — The Reference Guide
One card per model: provider, release date, context window, pricing, input and output modalities, how the model is marketed, and its RevenueBench characteristics (mean score, perfect passes, cost, fabrication count). Modalities in means what you can send the model; modalities out means what it can send back. Sorted by provider.
Coverage note: All 117 models ran the suite. Partial runs are labeled in the full matrix. Grok models are permanently excluded by decision.
anthropic 9 models
Anthropic: Claude Haiku 4.5
anthropic/claude-haiku-4.5
0.945 Released 2025-10-15 Context 200,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.945 mean · 30/40 perfect · 0 fabrications
Anthropic: Claude Opus 4.5
anthropic/claude-opus-4.5
0.966 Released 2025-11-24 Context 200,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.966 mean · 33/40 perfect · 0 fabrications
Anthropic: Claude Opus 4.7
anthropic/claude-opus-4.7
0.959 Released 2026-04-16 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.959 mean · 33/40 perfect · 0 fabrications
Anthropic: Claude Opus 4.8
anthropic/claude-opus-4.8
0.955 Released 2026-05-27 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.955 mean · 32/40 perfect · 0 fabrications
Anthropic: Claude Sonnet 5
anthropic/claude-sonnet-5
0.945 Released 2026-06-30 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.945 mean · 33/40 perfect · 0 fabrications
Anthropic: Claude Opus 5
anthropic/claude-opus-5
0.976 Released 2026-07-24 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.976 mean · 37/40 perfect · 0 fabrications
Anthropic: Claude Opus 5.5
anthropic/claude-opus-5.5
0.963 Released 2026-09-22 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.963 mean · 34/40 perfect · 0 fabrications
Anthropic: Claude Sonnet 5.5
anthropic/claude-sonnet-5.5
0.964 Released 2026-09-28 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.964 mean · 34/40 perfect · 0 fabrications
Anthropic: Claude Haiku 5.5
anthropic/claude-haiku-5.5
0.936 Released 2026-10-07 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Frontier reasoning & agentic work; marketed for enterprise safety, long-horizon coding, and nuanced writing.
RevenueBench 0.936 mean · 32/40 perfect · 0 fabrications
openai 29 models
Released 2025-04-16 Context 200,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.720 mean · 17/40 perfect · 0 fabrications
OpenAI: o4 Mini
openai/o4-mini
0.772 Released 2025-04-16 Context 200,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.772 mean · 17/40 perfect · 0 fabrications
OpenAI: GPT-5
openai/gpt-5
0.921 Released 2025-08-07 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.921 mean · 29/40 perfect · 1 fabrication
OpenAI: GPT-5 Mini
openai/gpt-5-mini
0.909 Released 2025-08-07 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.909 mean · 31/40 perfect · 0 fabrications
OpenAI: GPT-5 Nano
openai/gpt-5-nano
0.790 Released 2025-08-07 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.790 mean · 17/40 perfect · 0 fabrications
OpenAI: GPT-5 Pro
openai/gpt-5-pro
0.932 Released 2025-10-06 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.932 mean · 29/40 perfect · 0 fabrications
OpenAI: GPT-5.1
openai/gpt-5.1
0.940 Released 2025-11-13 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.940 mean · 30/40 perfect · 0 fabrications
OpenAI: GPT-5.2
openai/gpt-5.2
0.953 Released 2025-12-10 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.953 mean · 31/40 perfect · 0 fabrications
OpenAI: GPT-5.2 Pro
openai/gpt-5.2-pro
0.957 Released 2025-12-10 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.957 mean · 32/40 perfect · 0 fabrications
OpenAI: GPT-5.4 Pro
openai/gpt-5.4-pro
0.947 Released 2026-03-05 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.947 mean · 33/40 perfect · 0 fabrications
OpenAI: GPT-5.4 Mini
openai/gpt-5.4-mini
0.940 Released 2026-03-17 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.940 mean · 31/40 perfect · 1 fabrication
OpenAI: GPT-5.4 Nano
openai/gpt-5.4-nano
0.925 Released 2026-03-17 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.925 mean · 27/40 perfect · 0 fabrications
OpenAI: GPT-5.5
openai/gpt-5.5
0.959 Released 2026-04-24 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.959 mean · 33/40 perfect · 0 fabrications
OpenAI: GPT-5.5 Pro
openai/gpt-5.5-pro
0.966 Released 2026-04-24 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.966 mean · 34/40 perfect · 0 fabrications
OpenAI: GPT Chat Latest
openai/gpt-chat-latest
0.899 Released 2026-05-05 Context 400,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.899 mean · 27/40 perfect · 0 fabrications
OpenAI: GPT-5.6 Luna
openai/gpt-5.6-luna
0.941 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.941 mean · 31/40 perfect · 0 fabrications
OpenAI: GPT-5.6 Luna Pro
openai/gpt-5.6-luna-pro
0.939 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.939 mean · 30/40 perfect · 1 fabrication
OpenAI: GPT-5.6 Sol
openai/gpt-5.6-sol
0.945 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.945 mean · 30/40 perfect · 0 fabrications
OpenAI: GPT-5.6 Sol Pro
openai/gpt-5.6-sol-pro
0.942 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.942 mean · 32/40 perfect · 0 fabrications
OpenAI: GPT-5.6 Terra
openai/gpt-5.6-terra
0.925 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.925 mean · 28/40 perfect · 0 fabrications
OpenAI: GPT-5.6 Terra Pro
openai/gpt-5.6-terra-pro
0.920 Released 2026-07-09 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.920 mean · 27/40 perfect · 0 fabrications
OpenAI: GPT-6 Astra
openai/gpt-6-astra
0.957 Released 2026-09-04 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.957 mean · 35/40 perfect · 0 fabrications
OpenAI: GPT-6 Astra Pro
openai/gpt-6-astra-pro
0.946 Released 2026-09-04 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.946 mean · 33/40 perfect · 0 fabrications
OpenAI: GPT-6 Luna
openai/gpt-6-luna
0.910 Released 2026-09-22 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.910 mean · 28/40 perfect · 0 fabrications
OpenAI: GPT-6 Luna Pro
openai/gpt-6-luna-pro
0.879 Released 2026-09-22 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.879 mean · 26/40 perfect · 0 fabrications
OpenAI: GPT-6 Sol
openai/gpt-6-sol
0.946 Released 2026-09-22 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.946 mean · 30/40 perfect · 0 fabrications
OpenAI: GPT-6 Sol Pro
openai/gpt-6-sol-pro
0.926 Released 2026-09-22 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.926 mean · 30/40 perfect · 0 fabrications
OpenAI: GPT-6.1 Sol
openai/gpt-6.1-sol
0.945 Released 2026-09-29 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.945 mean · 34/40 perfect · 0 fabrications
OpenAI: GPT-6.1 Sol Pro
openai/gpt-6.1-sol-pro
0.955 Released 2026-09-29 Context 1,050,000 tok Modalities in: text, image, file · out: text Status 40/40
General-purpose frontier intelligence; marketed as the default choice for reasoning, coding, and agentic workflows.
RevenueBench 0.955 mean · 35/40 perfect · 0 fabrications
google 6 models
Google: Gemma 4 31B
google/gemma-4-31b-it
0.898 Released 2026-04-02 Context 262,144 tok Modalities in: text, image, video · out: text Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.898 mean · 26/40 perfect · 0 fabrications
Google: Gemma 4 26B A4B
google/gemma-4-26b-a4b-it
0.858 Released 2026-04-03 Context 262,144 tok Modalities in: text, image, video · out: text Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.858 mean · 22/40 perfect · 0 fabrications
Google: Nano Banana Pro (Gemini 3 Pro Image)
google/gemini-3-pro-image
0.873 Released 2026-06-18 Context 131,072 tok Modalities in: text, image · out: text, image Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.873 mean · 27/40 perfect · 1 fabrication
Google: Gemini 3.7 Flash
google/gemini-3.7-flash
0.962 Released 2026-08-13 Context 1,048,576 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.962 mean · 32/40 perfect · 0 fabrications
Google: Gemini 3.8 Flash
google/gemini-3.8-flash
0.980 Released 2026-09-02 Context 1,048,576 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.980 mean · 36/40 perfect · 0 fabrications
Google: Gemini 4 Argon
google/gemini-4-argon
0.965 Released 2026-10-13 Context 1,048,576 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Multimodal speed/cost efficiency at scale; marketed with massive context and integration across Google Cloud.
RevenueBench 0.965 mean · 34/40 perfect · 0 fabrications
amazon 2 models
Amazon: Nova Pro 1.0
amazon/nova-pro-v1
0.618 Released 2024-12-05 Context 300,000 tok Modalities in: text, image · out: text Status 40/40
Nova family; marketed for AWS-native enterprise deployment.
RevenueBench 0.618 mean · 9/40 perfect · 2 fabrications
Amazon: Nova Premier 1.0
amazon/nova-premier-v1
0.694 Released 2025-10-31 Context 1,000,000 tok Modalities in: text, image · out: text Status 40/40
Nova family; marketed for AWS-native enterprise deployment.
RevenueBench 0.694 mean · 16/40 perfect · 3 fabrications
Meta: Muse Spark 1.1
meta/muse-spark-1.1
0.986 Released 2026-07-16 Context 1,048,576 tok Modalities in: text, image, file, video · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.986 mean · 37/40 perfect · 0 fabrications
Meta: Muse Spark 1.2
meta/muse-spark-1.2
0.973 Released 2026-08-05 Context 1,048,576 tok Modalities in: text, image, file, video · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.973 mean · 35/40 perfect · 0 fabrications
Meta: Muse Glimmer 30B
meta/muse-glimmer-30b
0.938 Released 2026-08-09 Context 131,072 tok Modalities in: text, image · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.938 mean · 33/40 perfect · 0 fabrications
Meta: Muse Spark 1.3
meta/muse-spark-1.3
0.978 Released 2026-09-02 Context 1,048,576 tok Modalities in: text, image, file, video · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.978 mean · 36/40 perfect · 0 fabrications
Meta: Muse Wonder
meta/muse-wonder
0.996 Released 2026-10-12 Context 1,048,576 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.996 mean · 39/40 perfect · 0 fabrications
Meta: Muse Spark Open
meta/muse-spark-open
0.978 Released 2026-10-15 Context 1,048,576 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Meta's next-gen open-weights line; marketed for open research and on-prem deployment.
RevenueBench 0.978 mean · 36/40 perfect · 0 fabrications
moonshotai 2 models
MoonshotAI: Kimi K2.7 Code
moonshotai/kimi-k2.7-code
0.953 Released 2026-06-12 Context 262,144 tok Modalities in: text, image · out: text Status 40/40
Long-context specialists (Kimi); marketed on million-token context and strong agentic/coding results.
RevenueBench 0.953 mean · 31/40 perfect · 0 fabrications
MoonshotAI: Kimi K3
moonshotai/kimi-k3
0.985 Released 2026-07-16 Context 1,048,576 tok Modalities in: text, image, video · out: text Status 40/40
Long-context specialists (Kimi); marketed on million-token context and strong agentic/coding results.
RevenueBench 0.985 mean · 37/40 perfect · 0 fabrications
deepseek 7 models
DeepSeek: DeepSeek V4 Flash 0423
deepseek/deepseek-v4-flash
0.945 Released 2026-04-24 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.945 mean · 32/40 perfect · 1 fabrication
DeepSeek: DeepSeek V4 Pro 0423
deepseek/deepseek-v4-pro
0.964 Released 2026-04-24 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.964 mean · 34/40 perfect · 0 fabrications
DeepSeek: DeepSeek V4 Flash 0731
deepseek/deepseek-v4-flash-0731
0.953 Released 2026-07-31 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.953 mean · 32/40 perfect · 0 fabrications
DeepSeek: DeepSeek V4 Pro 0813
deepseek/deepseek-v4-pro-0813
0.952 Released 2026-08-12 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.952 mean · 33/40 perfect · 0 fabrications
DeepSeek: DeepSeek V4.1 Flash 1017
deepseek/deepseek-v4.1-flash-1017
0.956 Released 2026-10-17 Context 1,048,576 tok Modalities in: text, image · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.956 mean · 34/40 perfect · 0 fabrications
deepseek-flash-latest
deepseek/deepseek-flash-latest
0.965 Modalities in: text, image · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.965 mean · 34/40 perfect · 1 fabrication
deepseek-pro-latest
deepseek/deepseek-pro-latest
0.990 Modalities in: text · out: text Status 40/40
Ultra-cheap frontier-class reasoning; marketed on price-performance and open MoE weights.
RevenueBench 0.990 mean · 38/40 perfect · 0 fabrications
qwen 6 models
Qwen: Qwen3.8 2.4T A95B
qwen/qwen3.8-2.4t-a95b
0.965 Released 2026-08-12 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.965 mean · 34/40 perfect · 0 fabrications
Qwen: Qwen3.8 27B
qwen/qwen3.8-27b
0.962 Released 2026-08-14 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.962 mean · 34/40 perfect · 0 fabrications
Qwen: Qwen3.8 Flash
qwen/qwen3.8-flash
0.974 Released 2026-08-26 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.974 mean · 36/40 perfect · 0 fabrications
Qwen: Qwen3.8 Max (0902)
qwen/qwen3.8-max-0902
0.980 Released 2026-09-03 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.980 mean · 37/40 perfect · 0 fabrications
Qwen: Qwen3.8 Omni Flash
qwen/qwen3.8-omni-flash
0.970 Released 2026-09-21 Context 1,000,000 tok Modalities in: text, image, video, audio · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.970 mean · 34/40 perfect · 0 fabrications
Qwen: Qwen3.8 Max Prime
qwen/qwen3.8-max-prime
0.962 Released 2026-09-23 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
Alibaba's open-weight family spanning tiny to trillion-param; marketed on multilingual breadth and strong benchmarks per dollar.
RevenueBench 0.962 mean · 34/40 perfect · 0 fabrications
z-ai 7 models
Z.ai: GLM 5 Turbo
z-ai/glm-5-turbo
0.973 Released 2026-03-15 Context 202,752 tok Modalities in: text · out: text Status 31/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.973 mean · 28/31 perfect · 0 fabrications
Z.ai: GLM 5V Turbo
z-ai/glm-5v-turbo
0.982 Released 2026-04-01 Context 202,752 tok Modalities in: text, image, video · out: text Status 36/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.982 mean · 33/36 perfect · 1 fabrication
Z.ai: GLM 5.1
z-ai/glm-5.1
0.973 Released 2026-04-07 Context 204,800 tok Modalities in: text · out: text Status 40/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.973 mean · 34/40 perfect · 1 fabrication
Z.ai: GLM 5.3
z-ai/glm-5.3
0.970 Released 2026-08-18 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.970 mean · 34/40 perfect · 0 fabrications
Z.ai: GLM 5.3 Flash
z-ai/glm-5.3-flash
0.972 Released 2026-08-26 Context 1,048,576 tok Modalities in: text, image, video · out: text Status 40/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.972 mean · 35/40 perfect · 0 fabrications
Z.ai: GLM 5.3 FlashX
z-ai/glm-5.3-flashx
0.983 Released 2026-09-18 Context 1,048,576 tok Modalities in: text, image, video · out: text Status 40/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.983 mean · 37/40 perfect · 0 fabrications
Z.ai: GLM 5.3 Prime
z-ai/glm-5.3-prime
0.976 Released 2026-09-23 Context 1,000,000 tok Modalities in: text · out: text Status 37/40
Zhipu's GLM line; marketed as open-weight agentic models with strong tool-use.
RevenueBench 0.976 mean · 33/37 perfect · 0 fabrications
mistralai 1 model
Mistral: Mistral Large 4
mistralai/mistral-large-4-0
0.932 Released 2026-10-06 Context 1,048,576 tok Modalities in: text, image · out: text Status 40/40
European frontier-for-less; marketed on efficiency, open weights (smaller sizes), and EU data sovereignty.
RevenueBench 0.932 mean · 33/40 perfect · 1 fabrication
minimax 4 models
MiniMax: MiniMax M2
minimax/minimax-m2
0.919 Released 2025-10-23 Context 204,800 tok Modalities in: text · out: text Status 40/40
High-throughput commercial models; marketed for cost-efficient production serving.
RevenueBench 0.919 mean · 29/40 perfect · 0 fabrications
MiniMax: MiniMax M2.1
minimax/minimax-m2.1
0.942 Released 2025-12-23 Context 204,800 tok Modalities in: text · out: text Status 40/40
High-throughput commercial models; marketed for cost-efficient production serving.
RevenueBench 0.942 mean · 31/40 perfect · 0 fabrications
MiniMax: MiniMax M2.5
minimax/minimax-m2.5
0.924 Released 2026-02-12 Context 204,800 tok Modalities in: text · out: text Status 40/40
High-throughput commercial models; marketed for cost-efficient production serving.
RevenueBench 0.924 mean · 28/40 perfect · 0 fabrications
MiniMax: MiniMax M3.1 Flash (free)
minimax/minimax-m3.1-flash:free
0.953 Released 2026-10-19 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
High-throughput commercial models; marketed for cost-efficient production serving.
RevenueBench 0.953 mean · 32/40 perfect · 1 fabrication
tencent 3 models
Tencent: Hy3 preview
tencent/hy3-preview
0.977 Released 2026-04-22 Context 262,144 tok Modalities in: text · out: text Status 40/40
Hunyuan (Hy) models; marketed on strong Chinese+English reasoning at low cost.
RevenueBench 0.977 mean · 35/40 perfect · 0 fabrications
Released 2026-07-06 Context 262,144 tok Modalities in: text · out: text Status 40/40
Hunyuan (Hy) models; marketed on strong Chinese+English reasoning at low cost.
RevenueBench 0.967 mean · 36/40 perfect · 1 fabrication
Tencent: Hy4 preview
tencent/hy4-preview
0.911 Released 2026-08-28 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Hunyuan (Hy) models; marketed on strong Chinese+English reasoning at low cost.
RevenueBench 0.911 mean · 30/40 perfect · 0 fabrications
xiaomi 5 models
Xiaomi: MiMo-V2.5
xiaomi/mimo-v2.5
0.950 Released 2026-04-22 Context 1,050,000 tok Modalities in: text, image, video, audio · out: text Status 40/40
MiMo; marketed as efficient reasoning models from Xiaomi's phone-scale AI stack.
RevenueBench 0.950 mean · 33/40 perfect · 0 fabrications
Xiaomi: MiMo-V2.5-Pro
xiaomi/mimo-v2.5-pro
0.978 Released 2026-04-22 Context 1,050,000 tok Modalities in: text · out: text Status 40/40
MiMo; marketed as efficient reasoning models from Xiaomi's phone-scale AI stack.
RevenueBench 0.978 mean · 36/40 perfect · 0 fabrications
Xiaomi: MiMo-V2.6-Flash
xiaomi/mimo-v2.6-flash
0.968 Released 2026-09-21 Context 1,050,000 tok Modalities in: text, image, video, audio · out: text Status 36/40
MiMo; marketed as efficient reasoning models from Xiaomi's phone-scale AI stack.
RevenueBench 0.968 mean · 32/36 perfect · 0 fabrications
Xiaomi: MiMo-V2.6-Pro
xiaomi/mimo-v2.6-pro
0.989 Released 2026-09-21 Context 1,050,000 tok Modalities in: text, image, video, audio · out: text Status 38/40
MiMo; marketed as efficient reasoning models from Xiaomi's phone-scale AI stack.
RevenueBench 0.989 mean · 36/38 perfect · 0 fabrications
Xiaomi: MiMo-V2.6-Pro-UltraSpeed
xiaomi/mimo-v2.6-pro-ultraspeed
0.964 Released 2026-09-21 Context 1,048,576 tok Modalities in: text, image, video, audio · out: text Status 40/40
MiMo; marketed as efficient reasoning models from Xiaomi's phone-scale AI stack.
RevenueBench 0.964 mean · 35/40 perfect · 0 fabrications
nvidia 2 models
NVIDIA: Nemotron 3.5 Super VL
nvidia/nemotron-3.5-super-vl
0.945 Released 2026-10-15 Context 1,048,576 tok Modalities in: text, image, video · out: text Status 40/40
Nemotron line; marketed for enterprise fine-tuning on NVIDIA stack.
RevenueBench 0.945 mean · 31/40 perfect · 0 fabrications
NVIDIA: Nemotron 3.5 Super VL (free)
nvidia/nemotron-3.5-super-vl:free
0.949 Released 2026-10-15 Context 1,000,000 tok Modalities in: text, image, video · out: text Status 40/40
Nemotron line; marketed for enterprise fine-tuning on NVIDIA stack.
RevenueBench 0.949 mean · 33/40 perfect · 0 fabrications
ibm-granite 1 model
IBM: Granite 4.2 8B
ibm-granite/granite-4.2-8b
0.916 Released 2026-08-31 Context 131,072 tok Modalities in: text · out: text Status 39/40
Enterprise trust focus; marketed for transparent, IBM-supported business deployment.
RevenueBench 0.916 mean · 25/39 perfect · 0 fabrications
inclusionai 2 models
inclusionAI: Ling 3.0 Flash
inclusionai/ling-3.0-flash
0.924 Released 2026-07-23 Context 262,144 tok Modalities in: text · out: text Status 40/40
Ring/Ling; marketed as ultra-lightweight MoE for edge/cheap serving.
RevenueBench 0.924 mean · 29/40 perfect · 0 fabrications
inclusionAI: Ling 3.1 Flash
inclusionai/ling-3.1-flash
0.989 Released 2026-10-02 Context 262,144 tok Modalities in: text · out: text Status 23/40
Ring/Ling; marketed as ultra-lightweight MoE for edge/cheap serving.
RevenueBench 0.989 mean · 22/23 perfect · 0 fabrications
meituan 1 model
Meituan: LongCat 2.0
meituan/longcat-2.0
0.960 Released 2026-07-20 Context 1,048,756 tok Modalities in: text · out: text Status 40/40
LongCat; marketed for long-context open-weight deployment.
RevenueBench 0.960 mean · 33/40 perfect · 0 fabrications
nex-agi 2 models
nex-n2-mini
nex-agi/nex-n2-mini
0.946 Modalities in: text · out: text Status 28/40
Nex models; marketed for reasoning efficiency.
RevenueBench 0.946 mean · 23/28 perfect · 0 fabrications
nex-n2-pro
nex-agi/nex-n2-pro
0.929 Modalities in: text · out: text Status 14/40
Nex models; marketed for reasoning efficiency.
RevenueBench 0.929 mean · 10/14 perfect · 0 fabrications
poolside 1 model
Poolside: Laguna S 2.1
poolside/laguna-s-2.1
0.948 Released 2026-07-21 Context 1,048,576 tok Modalities in: text · out: text Status 39/40
Laguna; marketed for code analysis and software reasoning.
RevenueBench 0.948 mean · 33/39 perfect · 0 fabrications
sakana 1 model
Sakana: Fugu Ultra
sakana/fugu-ultra
0.962 Released 2026-06-24 Context 1,000,000 tok Modalities in: text, image · out: text Status 40/40
Japanese research lab models (fugu); marketed on novel training methods.
RevenueBench 0.962 mean · 35/40 perfect · 0 fabrications
stepfun 3 models
StepFun: Step 3.5 Flash
stepfun/step-3.5-flash
0.928 Released 2026-01-29 Context 262,144 tok Modalities in: text · out: text Status 40/40
Step; marketed for multimodal Chinese/English enterprise use.
RevenueBench 0.928 mean · 29/40 perfect · 1 fabrication
StepFun: Step 3.7 Flash
stepfun/step-3.7-flash
0.924 Released 2026-05-28 Context 262,144 tok Modalities in: text, image, video · out: text Status 40/40
Step; marketed for multimodal Chinese/English enterprise use.
RevenueBench 0.924 mean · 29/40 perfect · 1 fabrication
StepFun: Step 5 Preview
stepfun/step-5-preview
0.981 Released 2026-10-08 Context 1,048,576 tok Modalities in: text, image, video · out: text Status 40/40
Step; marketed for multimodal Chinese/English enterprise use.
RevenueBench 0.981 mean · 36/40 perfect · 0 fabrications
thinkingmachines 1 model
Thinking Machines: Inkling
thinkingmachines/inkling
0.921 Released 2026-07-17 Context 524,288 tok Modalities in: text, image, audio · out: text Status 40/40
Inkling; Thinking Machines Lab models.
RevenueBench 0.921 mean · 28/40 perfect · 0 fabrications
upstage 1 model
Upstage: Solar Pro 4
upstage/solar-pro4
0.918 Released 2026-08-10 Context 524,288 tok Modalities in: text · out: text Status 40/40
Solar; marketed for Korean/multilingual enterprise and document AI.
RevenueBench 0.918 mean · 30/40 perfect · 1 fabrication
aion-labs 2 models
AionLabs: Aion 3.5
aion-labs/aion-3.5
0.972 Released 2026-09-23 Context 262,144 tok Modalities in: text · out: text Status 33/40
Aion; newer entrant marketed on coding + reasoning balance.
RevenueBench 0.972 mean · 29/33 perfect · 0 fabrications
AionLabs: Aion 3.5 Mini
aion-labs/aion-3.5-mini
0.932 Released 2026-09-23 Context 262,144 tok Modalities in: text · out: text Status 36/40
Aion; newer entrant marketed on coding + reasoning balance.
RevenueBench 0.932 mean · 29/36 perfect · 0 fabrications
elorian 1 model
Elorian: Glimpse
elorian/glimpse
0.990 Released 2026-10-17 Context 262,144 tok Modalities in: text, image, video · out: text Status 40/40
Glimpse is a multimodal visual reasoning model from Elorian, built for visual understanding and complex visual reasoning
RevenueBench 0.990 mean · 38/40 perfect · 0 fabrications
fireworks 1 model
Fireworks: Ember-1
fireworks/ember-1
0.989 Released 2026-09-24 Context 1,048,576 tok Modalities in: text, image · out: text Status 40/40
Ember-1 is a specialized reasoning model from Fireworks Research, built on [Kimi K3](https://openrouter.ai/moonshotai/ki
RevenueBench 0.989 mean · 38/40 perfect · 0 fabrications
liquid 1 model
LiquidAI: LFM2.5-24B-A2B
liquid/lfm-2.5-24b-a2b
0.885 Released 2026-10-13 Context 128,000 tok Modalities in: text · out: text Status 34/40
Liquid Foundation Models; marketed on speed/quality frontier for edge.
RevenueBench 0.885 mean · 23/34 perfect · 0 fabrications
reflection 1 model
Reflection: Beam 1.0 501B A23B
reflection/beam-1.0-501b-a23b
0.914 Released 2026-10-16 Context 1,048,576 tok Modalities in: text · out: text Status 40/40
Beam 1.0 is Reflection's open-weight reasoning model for coding and long-horizon agentic workflows. It uses a sparse mix
RevenueBench 0.914 mean · 30/40 perfect · 0 fabrications
scaledown 2 models
ScaleDown: Extract
scaledown/extract
0.955 Released 2026-10-13 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Extract is a description-driven entity extraction model from ScaleDown. Instead of a fixed set of entity types, callers
RevenueBench 0.955 mean · 32/40 perfect · 0 fabrications
ScaleDown: Summarize
scaledown/summarize
0.981 Released 2026-10-13 Context 1,000,000 tok Modalities in: text, image, file · out: text Status 40/40
Summarize is an abstractive summarization model from ScaleDown. It condenses long text and documents into a fluent, fait
RevenueBench 0.981 mean · 36/40 perfect · 0 fabrications
stealth 1 model
space-bunny-alpha
stealth/space-bunny-alpha
0.947 Modalities in: text · out: text Status 38/40
stealth model.
RevenueBench 0.947 mean · 29/38 perfect · 0 fabrications
typesafe 1 model
TypeSafe: Jev Router
typesafe/jev-router
0.948 Released 2026-09-25 Context 1,000,000 tok Modalities in: text, image, file, video, audio · out: text Status 40/40
Jev Router picks the best model and reasoning effort for each request, balancing quality, speed, and cost. It runs on [J
RevenueBench 0.948 mean · 34/40 perfect · 0 fabrications
unbiased 4 models
Pareto 26.9
unbiased/pareto-26.9
0.982 Released 2026-10-13 Context 262,144 tok Modalities in: text, image · out: text Status 40/40
Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-leve
RevenueBench 0.982 mean · 37/40 perfect · 0 fabrications
Pareto Latest
~unbiased/pareto-latest
0.966 Released 2026-10-13 Context 1,048,576 tok Modalities in: text, image · out: text Status 40/40
This model always redirects to the latest model in the Pareto family.
RevenueBench 0.966 mean · 34/40 perfect · 0 fabrications
Pareto 26.10
unbiased/pareto-26.10
0.968 Released 2026-10-15 Context 1,048,576 tok Modalities in: text, image · out: text Status 40/40
Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-leve
RevenueBench 0.968 mean · 35/40 perfect · 1 fabrication
Modalities in: text · out: text Status 40/40
unbiased model.
RevenueBench 0.973 mean · 36/40 perfect · 0 fabrications
xingchen-agi 1 model
XingChen: Xing4.0 29B A4B (free)
xingchen-agi/xing4.0-29b-a4b:free
0.962 Released 2026-10-16 Context 262,144 tok Modalities in: text · out: text Status 40/40
Xing4.0-29B-A4B is an open-weight mixture-of-experts language model from China Telecom's AI team (TeleAI) and the succes
RevenueBench 0.962 mean · 34/40 perfect · 0 fabrications
RevenueBench model reference · 117 models · 33 providers run 2026-10-20 · deterministic scoring