Best small models in 2026: full ranking
Small models with the strongest benchmark averages. 159 models in the index carry a small-model tag or sit in the small open-weight category.
What the data shows
The best small model averages 66.2 and costs $2.02 per million input tokens. The group's median average score is 55.2.
Small models are usually the right default for classification, extraction and routing. They are less reliable on open-ended reasoning, so keep a larger model for the hard cases.
How to read the ranking
- Check the benchmark that matches your task.
- Compare blended price, not just input price.
- Test instruction-following on your prompts.
- Keep a larger fallback for edge cases.
Best small models
| # | Model | Provider | Input / 1M | Output / 1M | Context | Avg score |
|---|---|---|---|---|---|---|
| 1 | DeepSeek-V3.2 8B | DeepSeek | $2.02 | $9.37 | 32,768 | 66.2 |
| 2 | Kimi K2 8B | Moonshot AI | $1.86 | $10.03 | 8,192 | 64.7 |
| 3 | DeepSeek-R1 8B | DeepSeek | $1.87 | $8.53 | 16,384 | 64.3 |
| 4 | Llama 4 Maverick 8B | Meta AI | $1.64 | $8.15 | 4,096 | 63.4 |
| 5 | DeepSeek-V3.2 7B | DeepSeek | $1.75 | $6.84 | 4,096 | 63.2 |
| 6 | GLM-4.6 8B | Zhipu AI | $1.69 | $8.38 | 8,192 | 63.1 |
| 7 | DeepSeek-R1 7B | DeepSeek | $1.79 | $9.81 | 32,768 | 63.0 |
| 8 | Kimi K2 7B | Moonshot AI | $1.72 | $8.97 | 16,384 | 62.7 |
| 9 | Kimi K1.5 8B | Moonshot AI | $1.54 | $5.98 | 8,192 | 62.2 |
| 10 | Qwen3 8B | Alibaba Cloud | $1.66 | $6.65 | 8,192 | 61.9 |
| 11 | Llama 4 Maverick 7B | Meta AI | $1.52 | $8.19 | 4,096 | 61.9 |
| 12 | GLM-4.6 7B | Zhipu AI | $1.52 | $6.98 | 16,384 | 61.8 |
| 13 | MiniMax-M2 8B | MiniMax | $1.65 | $8.35 | 32,768 | 61.7 |
| 14 | Mistral Large 3 8B | Mistral AI | $1.62 | $6.82 | 16,384 | 61.6 |
| 15 | GLM-4.5 8B | Zhipu AI | $1.46 | $6.70 | 32,768 | 61.3 |
| 16 | Qwen3-Coder 8B | Alibaba Cloud | $1.43 | $7.30 | 32,768 | 61.3 |
| 17 | Llama 4 Scout 8B | Meta AI | $1.46 | $6.51 | 8,192 | 61.2 |
| 18 | Mistral Large 3 7B | Mistral AI | $1.44 | $7.37 | 32,768 | 61.1 |
| 19 | MiniMax-M2 7B | MiniMax | $1.44 | $7.00 | 8,192 | 60.8 |
| 20 | GLM-4.5 7B | Zhipu AI | $1.32 | $5.93 | 8,192 | 60.2 |
| 21 | Llama 3.3 8B | Meta AI | $1.26 | $5.56 | 8,192 | 60.1 |
| 22 | Llama 4 Scout 7B | Meta AI | $1.30 | $5.58 | 32,768 | 60.0 |
| 23 | Magistral 8B | Mistral AI | $1.33 | $6.03 | 4,096 | 60.0 |
| 24 | Kimi K1.5 7B | Moonshot AI | $1.35 | $4.98 | 8,192 | 59.9 |
| 25 | Qwen3 7B | Alibaba Cloud | $1.37 | $7.23 | 32,768 | 59.8 |
| 26 | Qwen3-Coder 7B | Alibaba Cloud | $1.24 | $5.31 | 32,768 | 59.6 |
| 27 | MiniMax-Text 8B | MiniMax | $1.22 | $5.08 | 4,096 | 59.5 |
| 28 | Llama 3.3 Groq 8B | Groq | $1.17 | $6.16 | 16,384 | 59.5 |
| 29 | QwQ 8B | Alibaba Cloud | $1.37 | $6.68 | 16,384 | 59.5 |
| 30 | Yi-Lightning 8B | 01.AI | $1.16 | $5.13 | 4,096 | 59.2 |
| 31 | Codestral 8B | Mistral AI | $1.18 | $5.35 | 4,096 | 59.0 |
| 32 | Magistral 7B | Mistral AI | $1.22 | $5.57 | 32,768 | 58.9 |
| 33 | Llama 3.1 Turbo 8B | Together AI | $1.16 | $5.48 | 32,768 | 58.8 |
| 34 | Llama 3.1 8B | Meta AI | $1.15 | $4.48 | 8,192 | 58.7 |
| 35 | DeepSeek-Coder 8B | DeepSeek | $1.26 | $6.25 | 32,768 | 58.7 |
| 36 | Hunyuan-Large 8B | Tencent | $1.29 | $6.63 | 4,096 | 58.7 |
| 37 | Llama 3.1 Cerebras 8B | Cerebras | $1.02 | $5.02 | 32,768 | 58.2 |
| 38 | QwQ 7B | Alibaba Cloud | $1.11 | $5.27 | 8,192 | 58.1 |
| 39 | Nemotron 4 8B | NVIDIA | $1.06 | $5.67 | 4,096 | 57.9 |
| 40 | Llama 3.3 7B | Meta AI | $1.18 | $5.59 | 8,192 | 57.8 |
| 41 | MiniMax-Text 7B | MiniMax | $1.10 | $4.16 | 16,384 | 57.7 |
| 42 | Hunyuan-Large 7B | Tencent | $1.08 | $5.56 | 32,768 | 57.7 |
| 43 | Command R+ 8B | Cohere | $1.21 | $5.69 | 8,192 | 57.7 |
| 44 | Llama 3.1 7B | Meta AI | $1.01 | $3.79 | 8,192 | 57.4 |
| 45 | Llama 3 DeepInfra 8B | DeepInfra | $1.05 | $3.99 | 8,192 | 57.4 |
| 46 | Llama 3 Anyscale 8B | Anyscale | $0.97 | $4.27 | 32,768 | 57.3 |
| 47 | DeepSeek-Coder 7B | DeepSeek | $1.14 | $4.37 | 32,768 | 57.3 |
| 48 | Llama 3 Lambda 8B | Lambda | $1.03 | $4.78 | 4,096 | 57.2 |
| 49 | Mistral Small 3 8B | Mistral AI | $0.99 | $5.38 | 8,192 | 57.2 |
| 50 | Seed-OSS 8B | ByteDance | $1.05 | $5.47 | 8,192 | 57.1 |
Methodology
This story uses the AI & LLM Index dataset: 600 models, 40 providers and 12 benchmarks. Prices are listed per million tokens. Scores are modelled estimates on a 0–100 scale and are not official benchmark results. Figures are rounded for readability.
Related
FAQ
What does the Best small models in 2026: full ranking data show?
Small models with the strongest benchmark averages. 159 models in the index carry a small-model tag or sit in the small open-weight category.
Are the numbers official?
No. Prices and scores are compiled for this demo index. Benchmark values are modelled estimates, not official results. Verify figures with the provider or benchmark owner before making decisions.
How should I use this ranking?
Use it to form a shortlist. Check the specific model pages, compare price and context, and test the shortlist on your own examples before committing.