Mistral · Model profile
Devstral Small 2
Intelligence, Performance & Price Analysis
Source: Canonical slug: devstral-small-2 · Canonical model registry at build time
Routing verdict
Capable everyday model
Good for routine tasks; route complex reasoning and premium workloads to stronger models.
Devstral Small 2 summary metrics
- Intelligence
- 17 scoreDevstral Small 2 scores 17 on the Artificial Analysis Intelligence Index, placing it well above average among other open weight non-reasoning models of similar size (median: 6).
- Speed
- 67.8 output tokens/secDevstral Small 2 generates output at 67.8 tokens per second (based on Mistral's API), which is at the lower end compared to other open weight non-reasoning models of similar size (median: 102.7 t/s).
- Latency
- 1.20s TTFTDevstral Small 2 has a time to first token (TTFT) of 1.20s (based on Mistral's API), which is better than average compared to other open weight non-reasoning models of similar size (median: 1.54s).
- Input Price
- Not published
- Output Price
- Not published
- Verbosity
- 9.5M Output tokens from Intelligence Index 2 out of 4 units for Verbosity . Comp
Assessment
Routing verdict and task fit
| Workload | Rating | Notes |
|---|---|---|
| Complex reasoning & agentic workflows | Viable | Intelligence score 17 handles routine reasoning but may struggle with open-ended agentic tasks. |
| High-volume chat & customer-facing | Optimal | Output speed 67.8 tokens/sec is adequate for chat. |
| Latency-sensitive applications | Optimal | TTFT 1.20s — adequate latency for most interactive use cases. |
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
- Llama 3.1 8B-30.3
- gpt-oss-120b (high)-50
- Mistral Small 3.1-50.3
- Mistral Small 3.2-50.4
- Qwen3.6 27B-52.8
- Devstral Small 2 (this model)-56.7
- Gemma 3 27B-65.9
- Ministral 3 14B-66.8
- Granite 4.1 30B-67.8
- Ministral 3 8B-67.8
| 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 |
| Llama 3.1 8B | -30.3 |
| gpt-oss-120b (high) | -50 |
| Mistral Small 3.1 | -50.3 |
| Mistral Small 3.2 | -50.4 |
| Qwen3.6 27B | -52.8 |
| Devstral Small 2 (this model) | -56.7 |
| Gemma 3 27B | -65.9 |
| Ministral 3 14B | -66.8 |
| Granite 4.1 30B | -67.8 |
| Ministral 3 8B | -67.8 |
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.
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
- Qwen3.6 27B30.5
- gpt-oss-120b (high)23.8
- Devstral Small 2 (this model)17.4
- Mistral Small 3.114.7
- Ministral 3 14B11.1
- Mistral Small 3.210.6
- Ministral 3 8B9
- Granite 4.1 30B8.9
- Llama 3.1 8B7.6
- Gemma 3 27B7.4
| 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 |
| Qwen3.6 27B | 30.5 |
| gpt-oss-120b (high) | 23.8 |
| Devstral Small 2 (this model) | 17.4 |
| Mistral Small 3.1 | 14.7 |
| Ministral 3 14B | 11.1 |
| Mistral Small 3.2 | 10.6 |
| Ministral 3 8B | 9 |
| Granite 4.1 30B | 8.9 |
| Llama 3.1 8B | 7.6 |
| Gemma 3 27B | 7.4 |
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
- Qwen3.6 27B30.5
- gpt-oss-120b (high)23.8
- Devstral Small 2 (this model)17.4
- Mistral Small 3.114.7
- Ministral 3 14B11.1
- Mistral Small 3.210.6
- Ministral 3 8B9
- Granite 4.1 30B8.9
- Llama 3.1 8B7.6
- Gemma 3 27B7.4
| 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 |
| Qwen3.6 27B | 30.5 |
| gpt-oss-120b (high) | 23.8 |
| Devstral Small 2 (this model) | 17.4 |
| Mistral Small 3.1 | 14.7 |
| Ministral 3 14B | 11.1 |
| Mistral Small 3.2 | 10.6 |
| Ministral 3 8B | 9 |
| Granite 4.1 30B | 8.9 |
| Llama 3.1 8B | 7.6 |
| Gemma 3 27B | 7.4 |
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
- Nemotron 3 Ultra83.3
- Granite 4.1 30B61.1
- DeepSeek V4 Pro (max)50
- Gemma 3 27B50
- Gemma 3 12B50
- GLM-5.2 (max)44.4
- gpt-oss-120b (high)38.9
- Qwen3.6 27B38.9
- Mistral Small 3.238.9
- Kimi K2.633.3
- MiniMax-M333.3
- EXAONE 4.5 33B27.8
| Model | Artificial Analysis Openness Index: Score |
|---|---|
| Nemotron 3 Ultra | 83.3 |
| Granite 4.1 30B | 61.1 |
| DeepSeek V4 Pro (max) | 50 |
| Gemma 3 27B | 50 |
| Gemma 3 12B | 50 |
| GLM-5.2 (max) | 44.4 |
| gpt-oss-120b (high) | 38.9 |
| Qwen3.6 27B | 38.9 |
| Mistral Small 3.2 | 38.9 |
| Kimi K2.6 | 33.3 |
| MiniMax-M3 | 33.3 |
| EXAONE 4.5 33B | 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
- Devstral Small 2 (this model)17.4
| 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 |
| Devstral Small 2 (this model) | 17.4 |
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
- gpt-oss-120b (high)316 t/s
- Nemotron 3 Ultra248 t/s
- GLM-5.2 (max)216 t/s
- Gemini 3.5 Flash192 t/s
- Mistral Small 3.1171 t/s
- Grok 4.3 (high)164 t/s
- Mistral Small 3.2147 t/s
- Llama 3.1 8B144 t/s
- MiniMax-M395 t/s
- Ministral 3 8B93 t/s
- GPT-5.5 (xhigh)88 t/s
- Ministral 3 14B78 t/s
- Kimi K2.676 t/s
- DeepSeek V4 Pro (max)72 t/s
- Claude Fable 5 (with fallback)71 t/s
- Devstral Small 2 (this model)68 t/s
- Qwen3.6 27B60 t/s
| Model | Output Speed |
|---|---|
| gpt-oss-120b (high) | 316 t/s |
| Nemotron 3 Ultra | 248 t/s |
| GLM-5.2 (max) | 216 t/s |
| Gemini 3.5 Flash | 192 t/s |
| Mistral Small 3.1 | 171 t/s |
| Grok 4.3 (high) | 164 t/s |
| Mistral Small 3.2 | 147 t/s |
| Llama 3.1 8B | 144 t/s |
| MiniMax-M3 | 95 t/s |
| Ministral 3 8B | 93 t/s |
| GPT-5.5 (xhigh) | 88 t/s |
| Ministral 3 14B | 78 t/s |
| Kimi K2.6 | 76 t/s |
| DeepSeek V4 Pro (max) | 72 t/s |
| Claude Fable 5 (with fallback) | 71 t/s |
| Devstral Small 2 (this model) | 68 t/s |
| Qwen3.6 27B | 60 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)316 t/s
- Nemotron 3 Ultra248 t/s
- GLM-5.2 (max)216 t/s
- Gemini 3.5 Flash192 t/s
- Grok 4.3 (high)164 t/s
- MiniMax-M395 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
- Devstral Small 2 (this model)68 t/s
| Model | Speed |
|---|---|
| gpt-oss-120b (high) | 316 t/s |
| Nemotron 3 Ultra | 248 t/s |
| GLM-5.2 (max) | 216 t/s |
| Gemini 3.5 Flash | 192 t/s |
| Grok 4.3 (high) | 164 t/s |
| MiniMax-M3 | 95 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 |
| Devstral Small 2 (this model) | 68 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
- Mistral Small 3.12.92 s
- Mistral Small 3.23.39 s
- Llama 3.1 8B3.48 s
- Ministral 3 8B5.35 s
- Ministral 3 14B6.4 s
- Devstral Small 2 (this model)7.37 s
- gpt-oss-120b (high)1.58 s
- Qwen3.6 27B8.39 s
- Nemotron 3 Ultra2.02 s
- GLM-5.2 (max)2.31 s
- Grok 4.3 (high)3.05 s
- MiniMax-M35.25 s
- Gemini 3.5 Flash2.6 s
- Kimi K2.66.56 s
- DeepSeek V4 Pro (max)6.95 s
- GPT-5.5 (xhigh)5.68 s
- Claude Fable 5 (with fallback)7.07 s
| Model | End-to-End Response Time |
|---|---|
| Mistral Small 3.1 | 2.92 s |
| Mistral Small 3.2 | 3.39 s |
| Llama 3.1 8B | 3.48 s |
| Ministral 3 8B | 5.35 s |
| Ministral 3 14B | 6.4 s |
| Devstral Small 2 (this model) | 7.37 s |
| gpt-oss-120b (high) | 1.58 s |
| Qwen3.6 27B | 8.39 s |
| Nemotron 3 Ultra | 2.02 s |
| GLM-5.2 (max) | 2.31 s |
| Grok 4.3 (high) | 3.05 s |
| MiniMax-M3 | 5.25 s |
| Gemini 3.5 Flash | 2.6 s |
| Kimi K2.6 | 6.56 s |
| DeepSeek V4 Pro (max) | 6.95 s |
| GPT-5.5 (xhigh) | 5.68 s |
| Claude Fable 5 (with fallback) | 7.07 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
- Mistral Small 3.20 s
- Mistral Small 3.10 s
- Ministral 3 8B0 s
- Ministral 3 14B0 s
- Llama 3.1 8B0 s
- Devstral Small 2 (this model)0 s
- Qwen3.6 27B0 s
- gpt-oss-120b (high)6.33 s
- Nemotron 3 Ultra9.17 s
- GLM-5.2 (max)9.26 s
- Grok 4.3 (high)0 s
- MiniMax-M320.98 s
- Gemini 3.5 Flash0 s
- Kimi K2.658.43 s
- DeepSeek V4 Pro (max)60.79 s
- GPT-5.5 (xhigh)0 s
- Claude Fable 5 (with fallback)0 s
| Model | Latency: Time To First Answer Token |
|---|---|
| Mistral Small 3.2 | 0 s |
| Mistral Small 3.1 | 0 s |
| Ministral 3 8B | 0 s |
| Ministral 3 14B | 0 s |
| Llama 3.1 8B | 0 s |
| Devstral Small 2 (this model) | 0 s |
| Qwen3.6 27B | 0 s |
| gpt-oss-120b (high) | 6.33 s |
| Nemotron 3 Ultra | 9.17 s |
| GLM-5.2 (max) | 9.26 s |
| Grok 4.3 (high) | 0 s |
| MiniMax-M3 | 20.98 s |
| Gemini 3.5 Flash | 0 s |
| Kimi K2.6 | 58.43 s |
| DeepSeek V4 Pro (max) | 60.79 s |
| GPT-5.5 (xhigh) | 0 s |
| Claude Fable 5 (with fallback) | 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
- 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
- GPT-5.5 (xhigh)922k
- Nemotron 3 Ultra262.1k
- Muse Spark262.1k
- Qwen3.6 27B262.1k
- EXAONE 4.5 33B262.1k
- Devstral Small 2 (this model)256k
- Kimi K2.6256k
- Ministral 3 14B256k
- Ministral 3 8B256k
- gpt-oss-120b (high)131.1k
- Granite 4.1 30B131.1k
- Mistral Small 3.1128k
- Mistral Small 3.2128k
- Llama 3.1 8B128k
| Model | Context Window |
|---|---|
| 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 |
| GPT-5.5 (xhigh) | 922k |
| Nemotron 3 Ultra | 262.1k |
| Muse Spark | 262.1k |
| Qwen3.6 27B | 262.1k |
| EXAONE 4.5 33B | 262.1k |
| Devstral Small 2 (this model) | 256k |
| Kimi K2.6 | 256k |
| Ministral 3 14B | 256k |
| Ministral 3 8B | 256k |
| gpt-oss-120b (high) | 131.1k |
| Granite 4.1 30B | 131.1k |
| Mistral Small 3.1 | 128k |
| Mistral Small 3.2 | 128k |
| Llama 3.1 8B | 128k |
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
- Mistral Small 3.10.00143
- DeepSeek V4 Pro (max)0.0063
- gpt-oss-120b (high)0.00454
- Mistral Small 3.20.00242
- MiniMax-M30.0139
- Grok 4.3 (high)0.0101
- Ministral 3 14B0.00235
- Ministral 3 8B0.00235
- Nemotron 3 Ultra0.0209
- Kimi K2.60.0216
- Qwen3.6 27B0.041
- GLM-5.2 (max)0.0268
- Gemini 3.5 Flash0.0884
- GPT-5.5 (xhigh)0.107
- Claude Fable 5 (with fallback)0.367
| Model | Cost per Intelligence Index Task |
|---|---|
| Mistral Small 3.1 | 0.00143 |
| DeepSeek V4 Pro (max) | 0.0063 |
| gpt-oss-120b (high) | 0.00454 |
| Mistral Small 3.2 | 0.00242 |
| MiniMax-M3 | 0.0139 |
| Grok 4.3 (high) | 0.0101 |
| Ministral 3 14B | 0.00235 |
| Ministral 3 8B | 0.00235 |
| Nemotron 3 Ultra | 0.0209 |
| Kimi K2.6 | 0.0216 |
| Qwen3.6 27B | 0.041 |
| GLM-5.2 (max) | 0.0268 |
| Gemini 3.5 Flash | 0.0884 |
| GPT-5.5 (xhigh) | 0.107 |
| 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
- Mistral Small 3.11.5
- gpt-oss-120b (high)3.5
- Ministral 3 8B2.8
- Ministral 3 14B3
- Mistral Small 3.22.5
- DeepSeek V4 Pro (max)7.2
- MiniMax-M314.5
- Grok 4.3 (high)12.5
- Qwen3.6 27B87.9
- Nemotron 3 Ultra33.5
- GLM-5.2 (max)32.9
- Kimi K2.625
- Gemini 3.5 Flash114
- GPT-5.5 (xhigh)131.8
- Claude Fable 5 (with fallback)507.6
| Model | Cost to Run Artificial Analysis Intelligence Index |
|---|---|
| Mistral Small 3.1 | 1.5 |
| gpt-oss-120b (high) | 3.5 |
| Ministral 3 8B | 2.8 |
| Ministral 3 14B | 3 |
| Mistral Small 3.2 | 2.5 |
| DeepSeek V4 Pro (max) | 7.2 |
| MiniMax-M3 | 14.5 |
| Grok 4.3 (high) | 12.5 |
| Qwen3.6 27B | 87.9 |
| Nemotron 3 Ultra | 33.5 |
| GLM-5.2 (max) | 32.9 |
| Kimi K2.6 | 25 |
| Gemini 3.5 Flash | 114 |
| 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
- Llama 3.1 8B$0.170
- Ministral 3 8B$0.300
- Mistral Small 3.1$0.400
- Ministral 3 14B$0.400
- Mistral Small 3.2$0.400
- gpt-oss-120b (high)$0.900
- DeepSeek V4 Pro (max)$1.31
- MiniMax-M3$1.56
- Nemotron 3 Ultra$3.60
- Grok 4.3 (high)$3.95
- Qwen3.6 27B$4.20
- Kimi K2.6$5.11
- GLM-5.2 (max)$6.06
- 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 |
|---|---|---|---|---|
| Llama 3.1 8B | $0.075 | $0.095 | Not published | $0.170 |
| Ministral 3 8B | $0.150 | $0.150 | Not published | $0.300 |
| Mistral Small 3.1 | $0.100 | $0.300 | Not published | $0.400 |
| Ministral 3 14B | $0.200 | $0.200 | Not published | $0.400 |
| Mistral Small 3.2 | $0.100 | $0.300 | Not published | $0.400 |
| gpt-oss-120b (high) | $0.150 | $0.600 | $0.150 | $0.900 |
| DeepSeek V4 Pro (max) | $0.435 | $0.870 | $0.0036 | $1.31 |
| MiniMax-M3 | $0.300 | $1.20 | $0.060 | $1.56 |
| Nemotron 3 Ultra | $0.675 | $2.67 | $0.250 | $3.60 |
| Grok 4.3 (high) | $1.25 | $2.50 | $0.200 | $3.95 |
| Qwen3.6 27B | $0.600 | $3.60 | Not published | $4.20 |
| Kimi K2.6 | $0.950 | $4.00 | $0.160 | $5.11 |
| GLM-5.2 (max) | $1.40 | $4.40 | $0.260 | $6.06 |
| 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
- Mistral Small 3.10.447
- Mistral Small 3.20.87
- Devstral Small 2 (this model)1.4
- Nemotron 3 Ultra1.4
- Grok 4.3 (high)1.6
- gpt-oss-120b (high)1.9
- Gemini 3.5 Flash2.2
- Ministral 3 14B2.2
- Ministral 3 8B2.5
- GLM-5.2 (max)2.8
- GPT-5.5 (xhigh)3
- Qwen3.6 27B3
- MiniMax-M34
- Claude Fable 5 (with fallback)4.8
- DeepSeek V4 Pro (max)6.7
- Kimi K2.68.2
| Model | Time per Intelligence Index Task |
|---|---|
| Mistral Small 3.1 | 0.447 |
| Mistral Small 3.2 | 0.87 |
| Devstral Small 2 (this model) | 1.4 |
| Nemotron 3 Ultra | 1.4 |
| Grok 4.3 (high) | 1.6 |
| gpt-oss-120b (high) | 1.9 |
| Gemini 3.5 Flash | 2.2 |
| Ministral 3 14B | 2.2 |
| Ministral 3 8B | 2.5 |
| GLM-5.2 (max) | 2.8 |
| GPT-5.5 (xhigh) | 3 |
| Qwen3.6 27B | 3 |
| MiniMax-M3 | 4 |
| Claude Fable 5 (with fallback) | 4.8 |
| DeepSeek V4 Pro (max) | 6.7 |
| Kimi K2.6 | 8.2 |
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.
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
- gpt-oss-120b (high)117B
- EXAONE 4.5 33B34.4B
- Granite 4.1 30B30B
- Qwen3.6 27B27.8B
- Gemma 3 27B27.4B
- Devstral Small 2 (this model)24B
- Mistral Small 3.124B
- Mistral Small 3.224B
- Ministral 3 14B14B
- Gemma 3 12B12.2B
- Ministral 3 8B8B
- Llama 3.1 8B8B
| 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 |
| gpt-oss-120b (high) | 111.9B | 5.1B | 117B |
| EXAONE 4.5 33B | 0B | 34.4B | 34.4B |
| Granite 4.1 30B | 0B | 30B | 30B |
| Qwen3.6 27B | 0B | 27.8B | 27.8B |
| Gemma 3 27B | 0B | 27.4B | 27.4B |
| Devstral Small 2 (this model) | 0B | 24B | 24B |
| Mistral Small 3.1 | 0B | 24B | 24B |
| Mistral Small 3.2 | 0B | 24B | 24B |
| Ministral 3 14B | 0B | 14B | 14B |
| Gemma 3 12B | 0B | 12.2B | 12.2B |
| Ministral 3 8B | 0B | 8B | 8B |
| Llama 3.1 8B | 0B | 8B | 8B |
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
- Granite 4.1 30B3,709.5
- Mistral Small 3.14,757.2
- Gemma 3 27B5,818.7
- Devstral Small 2 (this model)6,163.9
- Mistral Small 3.28,068
- Qwen3.6 27B11,390
- Ministral 3 14B11,745.9
- Muse Spark2,857.2
- Grok 4.3 (high)4,032
- Ministral 3 8B15,686.2
- Gemma 3 12B16,019.1
- GPT-5.5 (xhigh)3,572.5
- Nemotron 3 Ultra7,796.7
- MiniMax-M311,623.4
- Gemini 3.5 Flash9,819.4
- Claude Fable 5 (with fallback)7,695.8
- gpt-oss-120b (high)7,571.4
- DeepSeek V4 Pro (max)7,243.2
- Kimi K2.65,397.6
- GLM-5.2 (max)6,089.1
| Model | Output Tokens per Intelligence Index Task |
|---|---|
| Granite 4.1 30B | 3,709.5 |
| Mistral Small 3.1 | 4,757.2 |
| Gemma 3 27B | 5,818.7 |
| Devstral Small 2 (this model) | 6,163.9 |
| Mistral Small 3.2 | 8,068 |
| Qwen3.6 27B | 11,390 |
| Ministral 3 14B | 11,745.9 |
| Muse Spark | 2,857.2 |
| Grok 4.3 (high) | 4,032 |
| Ministral 3 8B | 15,686.2 |
| Gemma 3 12B | 16,019.1 |
| GPT-5.5 (xhigh) | 3,572.5 |
| Nemotron 3 Ultra | 7,796.7 |
| MiniMax-M3 | 11,623.4 |
| Gemini 3.5 Flash | 9,819.4 |
| Claude Fable 5 (with fallback) | 7,695.8 |
| gpt-oss-120b (high) | 7,571.4 |
| DeepSeek V4 Pro (max) | 7,243.2 |
| Kimi K2.6 | 5,397.6 |
| GLM-5.2 (max) | 6,089.1 |
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 | No | faq |
| Input modalities | Devstral Small 2 supports text and image input. | faq |
| Output modalities | Devstral Small 2 supports text output. | faq |
| Context window | 260k tokens | faq |
| Open weights / source | Yes, Devstral Small 2 is open weights. The model weights are publicly available and can be downloaded for self-hosting. | faq |
| Parameters | Devstral Small 2 has 24 billion parameters. | faq |
| License | Devstral Small 2 is released under the Apache 2.0 license. This license allows commercial use. | faq |
| API availability | Yes, Devstral Small 2 is available via API through 1 provider. | faq |
Method
Methodology & Provenance
This page is rendered from the normalized profile and page JSON for Devstral Small 2.
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.