Category ranking · Coding & agents
Best LLMs for Coding
Software engineering is where model differences are most visible. This ranking weights agentic coding, tool use, SWE-bench and terminal-task performance — the dimensions that matter for real codebases.
Benchmarks rotate and vendors use different harnesses. Treat this as a shortlist, then run your own evaluation. Data as of 2026-11-20.
| # | Model | Coding | Overall | Reasoning | Value | Price (in/out) | License |
|---|---|---|---|---|---|---|---|
| 1 | Anthropic · US | 99 | 93.3 | 97 | 70 | $4 / $20 | Proprietary |
| 2 | Moonshot AI · CN | 93 | 89.8 | 90 | 82 | $2 / $10 | Open weights |
| 3 | Anthropic · US | 92 | 88.9 | 87 | 90 | $2 / $10 | Proprietary |
| 4 | OpenAI · US | 91 | 90.1 | 96 | 62 | $10 / $50 | Proprietary |
| 5 | Z.ai (Zhipu) · CN | 89 | 84.5 | 83 | 90 | $0.60 / $2.20 | Open weights |
| 6 | OpenAI · US | 88 | 89.3 | 89 | 92 | $2 / $10 | Proprietary |
| 7 | DeepSeek · CN | 88 | 83.6 | 80 | 99 | $0.15 / $0.60 | Open source |
| 8 | xAI · US | 87 | 84.3 | 84 | 86 | $2 / $6 | Proprietary |
| 9 | Google · US | 86 | 89.7 | 88 | 84 | $2 / $10 | Proprietary |
| 10 | Alibaba Cloud · CN | 86 | 88.8 | 87 | 88 | $2 / $6 | Proprietary |
| 11 | Mistral AI · EU | 80 | 80.5 | 79 | 88 | $0.70 / $2.10 | Open weights |
| 12 | ByteDance · CN | 78 | 82.7 | 80 | 94 | $0.30 / $1.20 | Proprietary |
| 13 | Meta · US | 76 | 78.5 | 75 | 90 | $0.20 / $0.60 | Open source |