Best Coding Models
AI models ranked by code task performance. Scores reflect real success rates on coding requests routed through ANT.
Updated October 14, 2026
| Rank | Model | Provider | Code Success Rate | Avg Latency | Score |
|---|---|---|---|---|---|
| 1 | Llama 4 Scout | huggingface | 100.0% | 8007ms | 63.0 |
| 2 | DeepSeek V3.1 | huggingface | 100.0% | 6022ms | 63.0 |
| 3 | Gemini 2.5 Flash | huggingface | 100.0% | 2755ms | 57.0 |
| 4 | Qwen2.5 72B Instruct Heretic | featherless | 100.0% | 6778ms | 56.0 |
| 5 | Qwen3 Coder | nvidia | 100.0% | 12045ms | 51.0 |
| 6 | Qwen3.8 9B Distill Uncensored Heretic | featherless | 100.0% | 21112ms | 51.0 |
| 7 | o4-mini | openrouter | 100.0% | 21174ms | 48.0 |
| 8 | Qwen2.5 Coder 32b Instruct | cloudflare | 100.0% | 14870ms | 46.0 |
| 9 | o3 | openrouter | 100.0% | 21627ms | 45.0 |
| 10 | Claude Opus 4.6 | vercel | 100.0% | 140529ms | 45.0 |
| 11 | DeepHat V1 7B | unknown | 100.0% | 7ms | 40.0 |
| 12 | o3 | openai | 97.0% | 1500ms | 96.0 |
| 13 | OpenAI: GPT-5 Pro | openrouter | 96.0% | 2000ms | 97.0 |
| 14 | OpenAI: o1-pro | openrouter | 96.0% | 2000ms | 97.0 |
| 15 | GPT-5 | openai | 96.0% | 1000ms | 95.0 |
| 16 | GPT-5 | openai | 96.0% | 1000ms | 95.0 |
| 17 | OpenAI: GPT-4.1 | openrouter | 95.0% | 1000ms | 93.0 |
| 18 | GPT-4.1 | openai | 95.0% | 1000ms | 88.0 |
| 19 | Claude Sonnet 4 | anthropic | 94.0% | 1000ms | 92.0 |
| 20 | OpenAI: GPT-4o (2024-11-20) | openrouter | 93.0% | 1000ms | 92.0 |
| 21 | Anthropic: Claude 3.5 Sonnet | openrouter | 93.0% | 1500ms | 91.0 |
| 22 | Google: Gemini 2.5 Pro | openrouter | 92.0% | 1000ms | 94.0 |
| 23 | DeepSeek: R1 | openrouter | 92.0% | 600ms | 91.0 |
| 24 | o4-mini | openai | 91.0% | 1000ms | 90.0 |
| 25 | DeepSeek: DeepSeek V3.2 Speciale | openrouter | 91.0% | 600ms | 68.0 |
| 26 | DeepSeek: DeepSeek V3.2 Exp | openrouter | 91.0% | 600ms | 62.0 |
| 27 | Qwen: Qwen3 235B A22B | openrouter | 90.0% | 600ms | 84.0 |
| 28 | Qwen2.5 Coder 32B Instruct | openrouter | 90.0% | 600ms | 78.0 |
| 29 | Qwen: Qwen3 235B A22B Instruct 2507 | openrouter | 90.0% | 600ms | 57.0 |
| 30 | Qwen3 235B A22B | novita | 90.0% | 1000ms | 50.0 |
| 31 | Qwen3 235B A22B Instruct 2507 | novita | 90.0% | 600ms | 50.0 |
| 32 | Qwen3 235B A22b Thinking 2507 | novita | 90.0% | 1000ms | 50.0 |
| 33 | OpenAI: GPT-4 Turbo | openrouter | 89.0% | 1500ms | 88.0 |
| 34 | Google: Gemini 2.5 Flash | openrouter | 88.0% | 600ms | 85.0 |
| 35 | Google: Gemini 2.5 Flash Lite Preview 09-2025 | openrouter | 88.0% | 600ms | 66.0 |
| 36 | Google: Gemini 2.5 Flash Lite | openrouter | 88.0% | 600ms | 66.0 |
| 37 | OpenAI: o1 | openrouter | 87.0% | 2000ms | 95.0 |
| 38 | o1 | openai | 87.0% | 2000ms | 95.0 |
| 39 | o1 | openai | 87.0% | 2000ms | 95.0 |
| 40 | Mistral Large | openrouter | 87.0% | 1000ms | 73.0 |
| 41 | Mistral Large 2411 | openrouter | 87.0% | 1000ms | 73.0 |
| 42 | Mistral: Mistral Large 3 2512 | openrouter | 87.0% | 600ms | 72.0 |
| 43 | Mistral Large Pixtral 2411 | mistral | 87.0% | 400ms | 60.0 |
| 44 | Mistral Large 2411 | mistral | 87.0% | 400ms | 60.0 |
| 45 | Mistral Large 2512 | mistral | 87.0% | 400ms | 60.0 |
| 46 | Meta: Llama 3.3 70B Instruct | openrouter | 86.0% | 600ms | 80.0 |
| 47 | Meta: Llama 3.1 405B (base) | openrouter | 86.0% | 1000ms | 72.0 |
| 48 | Meta: Llama 3.1 405B Instruct | openrouter | 86.0% | 1000ms | 72.0 |
| 49 | Llama 3.3 70b Instruct Fp8 Fast | cloudflare | 86.0% | 400ms | 50.0 |
| 50 | Llama 3.3 70B Instruct | novita | 86.0% | 1000ms | 50.0 |