Mistral · Model profile
Mistral Small 4
Intelligence, Performance & Price Analysis
Source: Canonical slug: mistral-small-4 · Canonical model registry at build time
Routing verdict
Strong general-purpose model
Suitable for most production tasks, but high-volume or repetitive work should still be compared against cheaper routes.
Mistral Small 4 summary metrics
- Intelligence
- 20 scoreMistral Small 4 (Reasoning) scores 20 on the Artificial Analysis Intelligence Index, placing it well above average among other open weight models of similar size (median: 9).
- Speed
- 177.4 output tokens/secMistral Small 4 (Reasoning) generates output at 177.4 tokens per second (based on Mistral's API), which is well above average compared to other open weight models of similar size (median: 88.2 t/s).
- Latency
- 0.78s TTFTMistral Small 4 (Reasoning) has a time to first token (TTFT) of 0.78s (based on Mistral's API), which is very competitive compared to other open weight models of similar size (median: 1.52s).
- Input Price
- $0.15 / 1M tokensMistral Small 4 (Reasoning) costs $0.15 per 1M input tokens (very competitive, median: $0.40) and $0.60 per 1M output tokens (better than average, median: $0.84), based on Mistral's API.
- Output Price
- $0.60 / 1M tokensMistral Small 4 (Reasoning) costs $0.15 per 1M input tokens (very competitive, median: $0.40) and $0.60 per 1M output tokens (better than average, median: $0.84), based on Mistral's API.
- Verbosity
- 53M Output tokens from Intelligence Index 2 out of 4 units for Verbosity . Compa
Assessment
Routing verdict and task fit
| Workload | Rating | Notes |
|---|---|---|
| Complex reasoning & agentic workflows | Optimal | Intelligence score 20 supports capable reasoning, but very hard tasks may benefit from higher-tier models. |
| High-volume chat & customer-facing | Optimal | Output speed 177.4 tokens/sec and capable intelligence make this suitable for real-time chat at scale. |
| Latency-sensitive applications | Optimal | TTFT 0.78s — among the lowest latencies, suitable for interactive latency-critical use cases. |
| Cost-sensitive pipelines | Optimal | Output pricing at $0.60 is reasonable for moderate volume. |
Evidence
Evidence charts
AA-Omniscience Index
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis
- Claude Fable 5 (with fallback)40.1
- Gemini 3.5 Flash22.7
- GPT-5.5 (xhigh)20.1
- Grok 4.3 (high)18.3
- Kimi K2.66.4
- Muse Spark4.1
- GLM-5.2 (max)4
- MiniMax-M31.4
- Nemotron 3 Ultra-0.833
- DeepSeek V4 Pro (max)-10
- Mistral Small 4 (this model)-29.9
- K2 Think V2-33.9
- Mistral Medium 3.5-36.3
- Qwen3.5 122B A10B-39.6
- NVIDIA Nemotron 3 Super-42.1
- Devstral 2-46.2
- gpt-oss-120b (high)-50
- Qwen3 Next 80B A3B-50.6
- Llama 3.3 70B-51.9
- Llama 4 Scout-52.4
| Model | AA-Omniscience Index |
|---|---|
| Claude Fable 5 (with fallback) | 40.1 |
| Gemini 3.5 Flash | 22.7 |
| GPT-5.5 (xhigh) | 20.1 |
| Grok 4.3 (high) | 18.3 |
| Kimi K2.6 | 6.4 |
| Muse Spark | 4.1 |
| GLM-5.2 (max) | 4 |
| MiniMax-M3 | 1.4 |
| Nemotron 3 Ultra | -0.833 |
| DeepSeek V4 Pro (max) | -10 |
| Mistral Small 4 (this model) | -29.9 |
| K2 Think V2 | -33.9 |
| Mistral Medium 3.5 | -36.3 |
| Qwen3.5 122B A10B | -39.6 |
| NVIDIA Nemotron 3 Super | -42.1 |
| Devstral 2 | -46.2 |
| gpt-oss-120b (high) | -50 |
| Qwen3 Next 80B A3B | -50.6 |
| Llama 3.3 70B | -51.9 |
| Llama 4 Scout | -52.4 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Chart interpretation
- Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.
Artificial Analysis Intelligence Index by Open Weights / Proprietary
Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis
- Claude Fable 5 (with fallback)59.9
- GPT-5.5 (xhigh)54.8
- GLM-5.2 (max)51.1
- Gemini 3.5 Flash50.2
- MiniMax-M344.4
- DeepSeek V4 Pro (max)44.3
- Kimi K2.644.2
- Muse Spark43.1
- Nemotron 3 Ultra37.8
- Grok 4.3 (high)37.6
- Qwen3.5 122B A10B32.3
- Mistral Medium 3.529.9
- NVIDIA Nemotron 3 Super25.4
- gpt-oss-120b (high)23.8
- Qwen3 Coder Next21.1
- Mistral Small 4 (this model)19.6
- Devstral 219.2
- HyperNova 60B 260517.8
- K2 Think V217.3
- Qwen3 Next 80B A3B16.7
| Model | Artificial Analysis Intelligence Index by Open Weights / Proprietary |
|---|---|
| Claude Fable 5 (with fallback) | 59.9 |
| GPT-5.5 (xhigh) | 54.8 |
| GLM-5.2 (max) | 51.1 |
| Gemini 3.5 Flash | 50.2 |
| MiniMax-M3 | 44.4 |
| DeepSeek V4 Pro (max) | 44.3 |
| Kimi K2.6 | 44.2 |
| Muse Spark | 43.1 |
| Nemotron 3 Ultra | 37.8 |
| Grok 4.3 (high) | 37.6 |
| Qwen3.5 122B A10B | 32.3 |
| Mistral Medium 3.5 | 29.9 |
| NVIDIA Nemotron 3 Super | 25.4 |
| gpt-oss-120b (high) | 23.8 |
| Qwen3 Coder Next | 21.1 |
| Mistral Small 4 (this model) | 19.6 |
| Devstral 2 | 19.2 |
| HyperNova 60B 2605 | 17.8 |
| K2 Think V2 | 17.3 |
| Qwen3 Next 80B A3B | 16.7 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Artificial Analysis Intelligence Index
Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis
- Claude Fable 5 (with fallback)59.9
- GPT-5.5 (xhigh)54.8
- GLM-5.2 (max)51.1
- Gemini 3.5 Flash50.2
- MiniMax-M344.4
- DeepSeek V4 Pro (max)44.3
- Kimi K2.644.2
- Muse Spark43.1
- Nemotron 3 Ultra37.8
- Grok 4.3 (high)37.6
- Qwen3.5 122B A10B32.3
- Mistral Medium 3.529.9
- NVIDIA Nemotron 3 Super25.4
- gpt-oss-120b (high)23.8
- Qwen3 Coder Next21.1
- Mistral Small 4 (this model)19.6
- Devstral 219.2
- HyperNova 60B 260517.8
- K2 Think V217.3
- Qwen3 Next 80B A3B16.7
| Model | Artificial Analysis Intelligence Index |
|---|---|
| Claude Fable 5 (with fallback) | 59.9 |
| GPT-5.5 (xhigh) | 54.8 |
| GLM-5.2 (max) | 51.1 |
| Gemini 3.5 Flash | 50.2 |
| MiniMax-M3 | 44.4 |
| DeepSeek V4 Pro (max) | 44.3 |
| Kimi K2.6 | 44.2 |
| Muse Spark | 43.1 |
| Nemotron 3 Ultra | 37.8 |
| Grok 4.3 (high) | 37.6 |
| Qwen3.5 122B A10B | 32.3 |
| Mistral Medium 3.5 | 29.9 |
| NVIDIA Nemotron 3 Super | 25.4 |
| gpt-oss-120b (high) | 23.8 |
| Qwen3 Coder Next | 21.1 |
| Mistral Small 4 (this model) | 19.6 |
| Devstral 2 | 19.2 |
| HyperNova 60B 2605 | 17.8 |
| K2 Think V2 | 17.3 |
| Qwen3 Next 80B A3B | 16.7 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Artificial Analysis Openness Index: Score
Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open) · Evaluation results measured independently by Artificial Analysis
- K2 Think V288.9
- Nemotron 3 Ultra83.3
- NVIDIA Nemotron 3 Super83.3
- DeepSeek V4 Pro (max)50
- GLM-5.2 (max)44.4
- Qwen3 Next 80B A3B44.4
- Qwen3 Coder Next41.7
- Mistral Small 4 (this model)38.9
- gpt-oss-120b (high)38.9
- Qwen3.5 122B A10B38.9
- HyperNova 60B 260538.9
- Ling 2.6 Flash38.9
- Llama 3.3 70B38.9
- Kimi K2.633.3
- MiniMax-M333.3
- Mistral Medium 3.533.3
- Llama 4 Scout27.8
| Model | Artificial Analysis Openness Index: Score |
|---|---|
| K2 Think V2 | 88.9 |
| Nemotron 3 Ultra | 83.3 |
| NVIDIA Nemotron 3 Super | 83.3 |
| DeepSeek V4 Pro (max) | 50 |
| GLM-5.2 (max) | 44.4 |
| Qwen3 Next 80B A3B | 44.4 |
| Qwen3 Coder Next | 41.7 |
| Mistral Small 4 (this model) | 38.9 |
| gpt-oss-120b (high) | 38.9 |
| Qwen3.5 122B A10B | 38.9 |
| HyperNova 60B 2605 | 38.9 |
| Ling 2.6 Flash | 38.9 |
| Llama 3.3 70B | 38.9 |
| Kimi K2.6 | 33.3 |
| MiniMax-M3 | 33.3 |
| Mistral Medium 3.5 | 33.3 |
| Llama 4 Scout | 27.8 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Chart interpretation
- Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.
Intelligence
Artificial Analysis Intelligence Index · Higher is better · Evaluation results measured independently by Artificial Analysis
- Claude Fable 5 (with fallback)59.9
- GPT-5.5 (xhigh)54.8
- GLM-5.2 (max)51.1
- Gemini 3.5 Flash50.2
- MiniMax-M344.4
- DeepSeek V4 Pro (max)44.3
- Kimi K2.644.2
- Muse Spark43.1
- Nemotron 3 Ultra37.8
- Grok 4.3 (high)37.6
- gpt-oss-120b (high)23.8
- Mistral Small 4 (this model)19.6
| Model | Intelligence |
|---|---|
| Claude Fable 5 (with fallback) | 59.9 |
| GPT-5.5 (xhigh) | 54.8 |
| GLM-5.2 (max) | 51.1 |
| Gemini 3.5 Flash | 50.2 |
| MiniMax-M3 | 44.4 |
| DeepSeek V4 Pro (max) | 44.3 |
| Kimi K2.6 | 44.2 |
| Muse Spark | 43.1 |
| Nemotron 3 Ultra | 37.8 |
| Grok 4.3 (high) | 37.6 |
| gpt-oss-120b (high) | 23.8 |
| Mistral Small 4 (this model) | 19.6 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Output Speed
Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis
- NVIDIA Nemotron 3 Super364 t/s
- HyperNova 60B 2605353 t/s
- gpt-oss-120b (high)314 t/s
- Nemotron 3 Ultra250 t/s
- GLM-5.2 (max)218 t/s
- Gemini 3.5 Flash192 t/s
- Mistral Small 4 (this model)177 t/s
- Ling 2.6 Flash177 t/s
- Grok 4.3 (high)164 t/s
- Mistral Medium 3.5140 t/s
- Qwen3.5 122B A10B139 t/s
- Qwen3 Next 80B A3B115 t/s
- Llama 4 Scout109 t/s
- MiniMax-M399 t/s
- Qwen3 Coder Next89 t/s
- GPT-5.5 (xhigh)88 t/s
- Llama 3.3 70B86 t/s
- Devstral 276 t/s
- Kimi K2.676 t/s
- DeepSeek V4 Pro (max)72 t/s
| Model | Output Speed |
|---|---|
| NVIDIA Nemotron 3 Super | 364 t/s |
| HyperNova 60B 2605 | 353 t/s |
| gpt-oss-120b (high) | 314 t/s |
| Nemotron 3 Ultra | 250 t/s |
| GLM-5.2 (max) | 218 t/s |
| Gemini 3.5 Flash | 192 t/s |
| Mistral Small 4 (this model) | 177 t/s |
| Ling 2.6 Flash | 177 t/s |
| Grok 4.3 (high) | 164 t/s |
| Mistral Medium 3.5 | 140 t/s |
| Qwen3.5 122B A10B | 139 t/s |
| Qwen3 Next 80B A3B | 115 t/s |
| Llama 4 Scout | 109 t/s |
| MiniMax-M3 | 99 t/s |
| Qwen3 Coder Next | 89 t/s |
| GPT-5.5 (xhigh) | 88 t/s |
| Llama 3.3 70B | 86 t/s |
| Devstral 2 | 76 t/s |
| Kimi K2.6 | 76 t/s |
| DeepSeek V4 Pro (max) | 72 t/s |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Output speed context
- Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.
Speed
Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis
- gpt-oss-120b (high)314 t/s
- Nemotron 3 Ultra250 t/s
- GLM-5.2 (max)218 t/s
- Gemini 3.5 Flash192 t/s
- Mistral Small 4 (this model)177 t/s
- Grok 4.3 (high)164 t/s
- MiniMax-M399 t/s
- GPT-5.5 (xhigh)88 t/s
- Kimi K2.676 t/s
- DeepSeek V4 Pro (max)72 t/s
- Claude Fable 5 (with fallback)71 t/s
| Model | Speed |
|---|---|
| gpt-oss-120b (high) | 314 t/s |
| Nemotron 3 Ultra | 250 t/s |
| GLM-5.2 (max) | 218 t/s |
| Gemini 3.5 Flash | 192 t/s |
| Mistral Small 4 (this model) | 177 t/s |
| Grok 4.3 (high) | 164 t/s |
| MiniMax-M3 | 99 t/s |
| GPT-5.5 (xhigh) | 88 t/s |
| Kimi K2.6 | 76 t/s |
| DeepSeek V4 Pro (max) | 72 t/s |
| Claude Fable 5 (with fallback) | 71 t/s |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Output speed context
- Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.
End-to-End Response Time
Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better · Evaluation results measured independently by Artificial Analysis
- Ling 2.6 Flash2.82 s
- Llama 4 Scout4.61 s
- Llama 3.3 70B5.85 s
- Qwen3 Coder Next5.65 s
- Devstral 26.55 s
- HyperNova 60B 26051.42 s
- NVIDIA Nemotron 3 Super1.37 s
- gpt-oss-120b (high)1.59 s
- Nemotron 3 Ultra2 s
- GLM-5.2 (max)2.3 s
- Mistral Small 4 (this model)2.82 s
- Mistral Medium 3.53.56 s
- Qwen3.5 122B A10B3.59 s
- Qwen3 Next 80B A3B4.35 s
- Grok 4.3 (high)3.05 s
- MiniMax-M35.05 s
- Gemini 3.5 Flash2.6 s
- Kimi K2.66.56 s
- DeepSeek V4 Pro (max)6.97 s
- GPT-5.5 (xhigh)5.68 s
| Model | End-to-End Response Time |
|---|---|
| Ling 2.6 Flash | 2.82 s |
| Llama 4 Scout | 4.61 s |
| Llama 3.3 70B | 5.85 s |
| Qwen3 Coder Next | 5.65 s |
| Devstral 2 | 6.55 s |
| HyperNova 60B 2605 | 1.42 s |
| NVIDIA Nemotron 3 Super | 1.37 s |
| gpt-oss-120b (high) | 1.59 s |
| Nemotron 3 Ultra | 2 s |
| GLM-5.2 (max) | 2.3 s |
| Mistral Small 4 (this model) | 2.82 s |
| Mistral Medium 3.5 | 3.56 s |
| Qwen3.5 122B A10B | 3.59 s |
| Qwen3 Next 80B A3B | 4.35 s |
| Grok 4.3 (high) | 3.05 s |
| MiniMax-M3 | 5.05 s |
| Gemini 3.5 Flash | 2.6 s |
| Kimi K2.6 | 6.56 s |
| DeepSeek V4 Pro (max) | 6.97 s |
| GPT-5.5 (xhigh) | 5.68 s |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Chart interpretation
- Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.
Latency: Time To First Answer Token
Seconds to first answer token received · Accounts for reasoning model 'thinking' time · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout0 s
- Devstral 20 s
- Ling 2.6 Flash0 s
- Llama 3.3 70B0 s
- Qwen3 Coder Next0 s
- HyperNova 60B 26055.67 s
- NVIDIA Nemotron 3 Super5.5 s
- gpt-oss-120b (high)6.37 s
- Nemotron 3 Ultra9.09 s
- GLM-5.2 (max)9.19 s
- Mistral Small 4 (this model)11.27 s
- Mistral Medium 3.514.25 s
- Qwen3.5 122B A10B14.35 s
- Qwen3 Next 80B A3B17.41 s
- Grok 4.3 (high)0 s
- MiniMax-M320.19 s
- Gemini 3.5 Flash0 s
- Kimi K2.658.43 s
- DeepSeek V4 Pro (max)60.99 s
- GPT-5.5 (xhigh)0 s
| Model | Latency: Time To First Answer Token |
|---|---|
| Llama 4 Scout | 0 s |
| Devstral 2 | 0 s |
| Ling 2.6 Flash | 0 s |
| Llama 3.3 70B | 0 s |
| Qwen3 Coder Next | 0 s |
| HyperNova 60B 2605 | 5.67 s |
| NVIDIA Nemotron 3 Super | 5.5 s |
| gpt-oss-120b (high) | 6.37 s |
| Nemotron 3 Ultra | 9.09 s |
| GLM-5.2 (max) | 9.19 s |
| Mistral Small 4 (this model) | 11.27 s |
| Mistral Medium 3.5 | 14.25 s |
| Qwen3.5 122B A10B | 14.35 s |
| Qwen3 Next 80B A3B | 17.41 s |
| Grok 4.3 (high) | 0 s |
| MiniMax-M3 | 20.19 s |
| Gemini 3.5 Flash | 0 s |
| Kimi K2.6 | 58.43 s |
| DeepSeek V4 Pro (max) | 60.99 s |
| GPT-5.5 (xhigh) | 0 s |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Latency considerations
- Latency includes thinking / reasoning time when applicable. Lower is better for interactive use cases.
Context Window
Context window: tokens limit · Higher is better · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout10M
- Gemini 3.5 Flash1M
- Claude Fable 5 (with fallback)1M
- GLM-5.2 (max)1M
- DeepSeek V4 Pro (max)1M
- Grok 4.3 (high)1M
- MiniMax-M31M
- NVIDIA Nemotron 3 Super1M
- GPT-5.5 (xhigh)922k
- Nemotron 3 Ultra262.1k
- Muse Spark262.1k
- Qwen3.5 122B A10B262.1k
- K2 Think V2262.1k
- Qwen3 Next 80B A3B262.1k
- Ling 2.6 Flash262.1k
- Mistral Small 4 (this model)256k
- Kimi K2.6256k
- Mistral Medium 3.5256k
- Qwen3 Coder Next256k
- Devstral 2256k
| Model | Context Window |
|---|---|
| Llama 4 Scout | 10M |
| Gemini 3.5 Flash | 1M |
| Claude Fable 5 (with fallback) | 1M |
| GLM-5.2 (max) | 1M |
| DeepSeek V4 Pro (max) | 1M |
| Grok 4.3 (high) | 1M |
| MiniMax-M3 | 1M |
| NVIDIA Nemotron 3 Super | 1M |
| GPT-5.5 (xhigh) | 922k |
| Nemotron 3 Ultra | 262.1k |
| Muse Spark | 262.1k |
| Qwen3.5 122B A10B | 262.1k |
| K2 Think V2 | 262.1k |
| Qwen3 Next 80B A3B | 262.1k |
| Ling 2.6 Flash | 262.1k |
| Mistral Small 4 (this model) | 256k |
| Kimi K2.6 | 256k |
| Mistral Medium 3.5 | 256k |
| Qwen3 Coder Next | 256k |
| Devstral 2 | 256k |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Context window sizing
- Larger context windows enable complex multi-turn tasks but may increase latency and cost for long inputs.
Cost per Intelligence Index Task
Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout0.000824
- HyperNova 60B 26050.000717
- DeepSeek V4 Pro (max)0.0063
- gpt-oss-120b (high)0.00454
- Llama 3.3 70B0.0012
- MiniMax-M30.0139
- Grok 4.3 (high)0.0101
- Qwen3 Next 80B A3B0.0111
- NVIDIA Nemotron 3 Super0.00577
- Qwen3.5 122B A10B0.0208
- Nemotron 3 Ultra0.0209
- Qwen3 Coder Next0.0184
- Kimi K2.60.0216
- GLM-5.2 (max)0.0268
- Gemini 3.5 Flash0.0884
- GPT-5.5 (xhigh)0.107
- Mistral Medium 3.50.0529
- Claude Fable 5 (with fallback)0.367
| Model | Cost per Intelligence Index Task |
|---|---|
| Llama 4 Scout | 0.000824 |
| HyperNova 60B 2605 | 0.000717 |
| DeepSeek V4 Pro (max) | 0.0063 |
| gpt-oss-120b (high) | 0.00454 |
| Llama 3.3 70B | 0.0012 |
| MiniMax-M3 | 0.0139 |
| Grok 4.3 (high) | 0.0101 |
| Qwen3 Next 80B A3B | 0.0111 |
| NVIDIA Nemotron 3 Super | 0.00577 |
| Qwen3.5 122B A10B | 0.0208 |
| Nemotron 3 Ultra | 0.0209 |
| Qwen3 Coder Next | 0.0184 |
| Kimi K2.6 | 0.0216 |
| GLM-5.2 (max) | 0.0268 |
| Gemini 3.5 Flash | 0.0884 |
| GPT-5.5 (xhigh) | 0.107 |
| Mistral Medium 3.5 | 0.0529 |
| Claude Fable 5 (with fallback) | 0.367 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
- Cost pressure
- Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.
Cost per Task
Weighted average cost (USD) per Intelligence Index task · Lower is better · Evaluation results measured independently by Artificial Analysis
- DeepSeek V4 Pro (max)$0.045
- gpt-oss-120b (high)$0.061
- MiniMax-M3$0.125
- Grok 4.3 (high)$0.139
- Nemotron 3 Ultra$0.244
- Kimi K2.6$0.346
- GLM-5.2 (max)$0.370
- Gemini 3.5 Flash$0.586
- GPT-5.5 (xhigh)$0.863
- Claude Fable 5 (with fallback)$2.75
| Model | Cost per Task |
|---|---|
| DeepSeek V4 Pro (max) | $0.045 |
| gpt-oss-120b (high) | $0.061 |
| MiniMax-M3 | $0.125 |
| Grok 4.3 (high) | $0.139 |
| Nemotron 3 Ultra | $0.244 |
| Kimi K2.6 | $0.346 |
| GLM-5.2 (max) | $0.370 |
| Gemini 3.5 Flash | $0.586 |
| GPT-5.5 (xhigh) | $0.863 |
| Claude Fable 5 (with fallback) | $2.75 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Cost pressure
- Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.
Cost to Run Artificial Analysis Intelligence Index
Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout2.7
- HyperNova 60B 26050.947
- Llama 3.3 70B2.7
- gpt-oss-120b (high)3.5
- DeepSeek V4 Pro (max)7.2
- MiniMax-M314.5
- Qwen3 Coder Next38.5
- NVIDIA Nemotron 3 Super5.4
- Grok 4.3 (high)12.5
- Qwen3 Next 80B A3B15.9
- Nemotron 3 Ultra33.5
- Qwen3.5 122B A10B32.7
- GLM-5.2 (max)32.9
- Kimi K2.625
- Gemini 3.5 Flash114
- Mistral Medium 3.553.7
- GPT-5.5 (xhigh)131.8
- Claude Fable 5 (with fallback)507.6
| Model | Cost to Run Artificial Analysis Intelligence Index |
|---|---|
| Llama 4 Scout | 2.7 |
| HyperNova 60B 2605 | 0.947 |
| Llama 3.3 70B | 2.7 |
| gpt-oss-120b (high) | 3.5 |
| DeepSeek V4 Pro (max) | 7.2 |
| MiniMax-M3 | 14.5 |
| Qwen3 Coder Next | 38.5 |
| NVIDIA Nemotron 3 Super | 5.4 |
| Grok 4.3 (high) | 12.5 |
| Qwen3 Next 80B A3B | 15.9 |
| Nemotron 3 Ultra | 33.5 |
| Qwen3.5 122B A10B | 32.7 |
| GLM-5.2 (max) | 32.9 |
| Kimi K2.6 | 25 |
| Gemini 3.5 Flash | 114 |
| Mistral Medium 3.5 | 53.7 |
| GPT-5.5 (xhigh) | 131.8 |
| Claude Fable 5 (with fallback) | 507.6 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
- Cost pressure
- Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.
Pricing: Cache Hit, Input, and Output
Price (USD per M Tokens) · Evaluation results measured independently by Artificial Analysis
- Input
- Output
- Cache Hit
- HyperNova 60B 2605$0.180
- Ling 2.6 Flash$0.420
- Mistral Small 4 (this model)$0.750
- gpt-oss-120b (high)$0.900
- Llama 4 Scout$0.830
- NVIDIA Nemotron 3 Super$1.17
- Llama 3.3 70B$1.87
- DeepSeek V4 Pro (max)$1.31
- MiniMax-M3$1.56
- Qwen3 Coder Next$1.90
- Nemotron 3 Ultra$3.60
- Qwen3.5 122B A10B$3.60
- Grok 4.3 (high)$3.95
- Kimi K2.6$5.11
- GLM-5.2 (max)$6.06
- Qwen3 Next 80B A3B$6.50
- Mistral Medium 3.5$9.15
- Gemini 3.5 Flash$10.65
- GPT-5.5 (xhigh)$35.50
- Claude Fable 5 (with fallback)$61.00
| Model | Input | Output | Cache Hit | Total |
|---|---|---|---|---|
| HyperNova 60B 2605 | $0.040 | $0.140 | Not published | $0.180 |
| Ling 2.6 Flash | $0.100 | $0.300 | $0.020 | $0.420 |
| Mistral Small 4 (this model) | $0.150 | $0.600 | Not published | $0.750 |
| gpt-oss-120b (high) | $0.150 | $0.600 | $0.150 | $0.900 |
| Llama 4 Scout | $0.170 | $0.660 | Not published | $0.830 |
| NVIDIA Nemotron 3 Super | $0.250 | $0.775 | $0.143 | $1.17 |
| Llama 3.3 70B | $0.580 | $0.710 | $0.580 | $1.87 |
| DeepSeek V4 Pro (max) | $0.435 | $0.870 | $0.0036 | $1.31 |
| MiniMax-M3 | $0.300 | $1.20 | $0.060 | $1.56 |
| Qwen3 Coder Next | $0.350 | $1.20 | $0.350 | $1.90 |
| Nemotron 3 Ultra | $0.675 | $2.67 | $0.250 | $3.60 |
| Qwen3.5 122B A10B | $0.400 | $3.20 | Not published | $3.60 |
| Grok 4.3 (high) | $1.25 | $2.50 | $0.200 | $3.95 |
| Kimi K2.6 | $0.950 | $4.00 | $0.160 | $5.11 |
| GLM-5.2 (max) | $1.40 | $4.40 | $0.260 | $6.06 |
| Qwen3 Next 80B A3B | $0.500 | $6.00 | Not published | $6.50 |
| Mistral Medium 3.5 | $1.50 | $7.50 | $0.150 | $9.15 |
| Gemini 3.5 Flash | $1.50 | $9.00 | $0.150 | $10.65 |
| GPT-5.5 (xhigh) | $5.00 | $30.00 | $0.500 | $35.50 |
| Claude Fable 5 (with fallback) | $10.00 | $50.00 | $1.00 | $61.00 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Chart interpretation
- Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.
Time per Intelligence Index Task
Weighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout0.198
- Llama 3.3 70B0.321
- NVIDIA Nemotron 3 Super1.1
- Ling 2.6 Flash1.1
- HyperNova 60B 26051.2
- Devstral 21.3
- Nemotron 3 Ultra1.4
- Grok 4.3 (high)1.6
- Mistral Small 4 (this model)1.7
- Qwen3.5 122B A10B1.8
- gpt-oss-120b (high)1.9
- Gemini 3.5 Flash2.2
- Qwen3 Coder Next2.6
- Qwen3 Next 80B A3B2.8
- GLM-5.2 (max)2.8
- Mistral Medium 3.52.9
- GPT-5.5 (xhigh)3
- MiniMax-M33.9
- Claude Fable 5 (with fallback)4.8
- DeepSeek V4 Pro (max)6.8
| Model | Time per Intelligence Index Task |
|---|---|
| Llama 4 Scout | 0.198 |
| Llama 3.3 70B | 0.321 |
| NVIDIA Nemotron 3 Super | 1.1 |
| Ling 2.6 Flash | 1.1 |
| HyperNova 60B 2605 | 1.2 |
| Devstral 2 | 1.3 |
| Nemotron 3 Ultra | 1.4 |
| Grok 4.3 (high) | 1.6 |
| Mistral Small 4 (this model) | 1.7 |
| Qwen3.5 122B A10B | 1.8 |
| gpt-oss-120b (high) | 1.9 |
| Gemini 3.5 Flash | 2.2 |
| Qwen3 Coder Next | 2.6 |
| Qwen3 Next 80B A3B | 2.8 |
| GLM-5.2 (max) | 2.8 |
| Mistral Medium 3.5 | 2.9 |
| GPT-5.5 (xhigh) | 3 |
| MiniMax-M3 | 3.9 |
| Claude Fable 5 (with fallback) | 4.8 |
| DeepSeek V4 Pro (max) | 6.8 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Model Size: Total and Active Parameters
Comparison between total model parameters and parameters active during inference · Evaluation results measured independently by Artificial Analysis
- Passive Params
- Active Params
- DeepSeek V4 Pro (max)1,600B
- Kimi K2.61,000B
- GLM-5.2 (max)753B
- Nemotron 3 Ultra550B
- MiniMax-M3428B
- Mistral Medium 3.5128B
- Qwen3.5 122B A10B125B
- Devstral 2125B
- NVIDIA Nemotron 3 Super120.6B
- Mistral Small 4 (this model)119B
- gpt-oss-120b (high)117B
- Llama 4 Scout109B
- Ling 2.6 Flash107B
- Qwen3 Next 80B A3B80B
- Qwen3 Coder Next79.7B
- K2 Think V270B
- Llama 3.3 70B70B
- HyperNova 60B 260558.7B
| Model | Passive Params | Active Params | Total |
|---|---|---|---|
| DeepSeek V4 Pro (max) | 1,551B | 49B | 1,600B |
| Kimi K2.6 | 968B | 32B | 1,000B |
| GLM-5.2 (max) | 713B | 40B | 753B |
| Nemotron 3 Ultra | 495B | 55B | 550B |
| MiniMax-M3 | 405B | 23B | 428B |
| Mistral Medium 3.5 | 0B | 128B | 128B |
| Qwen3.5 122B A10B | 115B | 10B | 125B |
| Devstral 2 | 0B | 125B | 125B |
| NVIDIA Nemotron 3 Super | 107.9B | 12.7B | 120.6B |
| Mistral Small 4 (this model) | 112.5B | 6.5B | 119B |
| gpt-oss-120b (high) | 111.9B | 5.1B | 117B |
| Llama 4 Scout | 92B | 17B | 109B |
| Ling 2.6 Flash | 99.6B | 7.4B | 107B |
| Qwen3 Next 80B A3B | 77B | 3B | 80B |
| Qwen3 Coder Next | 76.7B | 3B | 79.7B |
| K2 Think V2 | 0B | 70B | 70B |
| Llama 3.3 70B | 0B | 70B | 70B |
| HyperNova 60B 2605 | 53.9B | 4.8B | 58.7B |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Architecture scale
- Larger models generally perform better but cost more to serve. Active parameter count matters for inference efficiency.
Output Tokens per Intelligence Index Task
Weighted average number of output tokens used to run one task in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis
- Llama 4 Scout1,249.2
- Llama 3.3 70B1,685.9
- Devstral 25,946.4
- Muse Spark2,857.2
- Ling 2.6 Flash13,070.1
- Grok 4.3 (high)4,032
- Qwen3 Coder Next15,355.4
- Mistral Small 4 (this model)6,364.2
- GPT-5.5 (xhigh)3,572.5
- Qwen3.5 122B A10B6,491.8
- Qwen3 Next 80B A3B1,857.4
- Nemotron 3 Ultra7,796.7
- MiniMax-M311,623.4
- Mistral Medium 3.57,048.7
- NVIDIA Nemotron 3 Super7,442.1
- Gemini 3.5 Flash9,819.4
- HyperNova 60B 26055,120
- Claude Fable 5 (with fallback)7,695.8
- gpt-oss-120b (high)7,571.4
- DeepSeek V4 Pro (max)7,243.2
| Model | Output Tokens per Intelligence Index Task |
|---|---|
| Llama 4 Scout | 1,249.2 |
| Llama 3.3 70B | 1,685.9 |
| Devstral 2 | 5,946.4 |
| Muse Spark | 2,857.2 |
| Ling 2.6 Flash | 13,070.1 |
| Grok 4.3 (high) | 4,032 |
| Qwen3 Coder Next | 15,355.4 |
| Mistral Small 4 (this model) | 6,364.2 |
| GPT-5.5 (xhigh) | 3,572.5 |
| Qwen3.5 122B A10B | 6,491.8 |
| Qwen3 Next 80B A3B | 1,857.4 |
| Nemotron 3 Ultra | 7,796.7 |
| MiniMax-M3 | 11,623.4 |
| Mistral Medium 3.5 | 7,048.7 |
| NVIDIA Nemotron 3 Super | 7,442.1 |
| Gemini 3.5 Flash | 9,819.4 |
| HyperNova 60B 2605 | 5,120 |
| Claude Fable 5 (with fallback) | 7,695.8 |
| gpt-oss-120b (high) | 7,571.4 |
| DeepSeek V4 Pro (max) | 7,243.2 |
Published benchmark dataset · Retrieved
Method and source
- Methodology
- Independent test run by Artificial Analysis on dedicated hardware.
- Benchmark interpretation
- Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Specifications
Architecture and limits
| Specification | Value | Source |
|---|---|---|
| Model type | Open weights | faq |
| Reasoning | Yes | faq |
| Input modalities | Mistral Small 4 (Reasoning) supports text and image input. | faq |
| Output modalities | Mistral Small 4 (Reasoning) supports text output. | faq |
| Context window | 260k tokens | faq |
| Open weights / source | Yes, Mistral Small 4 (Reasoning) is open weights. The model weights are publicly available and can be downloaded for self-hosting. | faq |
| Parameters | Mistral Small 4 (Reasoning) has 119 billion parameters (6.5 billion active). | faq |
| Active parameters | Mistral Small 4 (Reasoning) is a Mixture of Experts (MoE) model with 119 billion total parameters, but only 6.5 billion active parameters are used during inference. | faq |
| License | Mistral Small 4 (Reasoning) is released under the Apache 2.0 license. This license allows commercial use. | faq |
| API availability | Yes, Mistral Small 4 (Reasoning) is available via API through 1 provider. | faq |
Method
Methodology & Provenance
This page is rendered from the normalized profile and page JSON for Mistral Small 4.
Benchmark values are preserved as normalized; only layout, disclosure ordering, and typography are adjusted for readability.
Questions
Model FAQ and technical disclosures
Models
Related
- 01MistralOther Mistral models in the catalog.
- 02LeaderboardsCertified metric ranks, when published.