Best LLMs for coding
Ranked by Design Arena's code category — third-party, head-to-head preference votes on coding tasks. 94models in the catalog carry this benchmark; the top 30 are below. These are benchmark rankings, not usage rankings: OpenKey's own usage data accumulates from the July 2026 launch and will be added here, whale-filtered, once it exists.
Looking for the catalog view instead? Browse coding models — or the overall rankings.
Design Arena · code category elo · catalog snapshot Jul 1, 2026
| # | Model | Elo | Win rate | Input /1M | Output /1M | Context |
|---|---|---|---|---|---|---|
| 1 |  GLM 5.2by Z.ai | 1360 | 62% | $0.958 | $3.09 | 1.0M |
| 2 |  Claude Fable 5by Anthropic | 1349 | 63.3% | $10.30 | $51.50 | 1M |
| 3 | Claude Opus 4.6by Anthropic | 1342 | 63.9% | $5.15 | $25.75 | 1M |
| 4 | Claude Opus 4.7by Anthropic | 1337 | 61.1% | $5.15 | $25.75 | 1M |
| 5 | GLM 5.1by Z.ai | 1332 | 60.3% | $1.00 | $4.43 | 203K |
| 6 | Claude Sonnet 4.6by Anthropic | 1326 | 61.5% | $3.09 | $15.45 | 1M |
| 7 |  Kimi K2.6by Moonshot AI | 1326 | 58.1% | $0.567 | $3.30 | 262K |
| 8 | GLM 5 Turboby Z.ai | 1320 | 57.6% | $1.24 | $4.12 | 262K |
| 9 |  MiMo-V2.5-Proby Xiaomi | 1318 | 58.1% | $0.448 | $0.896 | 1.0M |
| 10 |  Qwen3.7 Maxby Qwen | 1311 | 57.5% | $1.29 | $3.86 | 1M |
| 11 | Kimi K2.7 Codeby Moonshot AI | 1307 | 55.6% | $0.762 | $3.60 | 262K |
| 12 |  Gemini 3.5 Flashby Google | 1306 | 58.5% | $1.54 | $9.27 | 1.0M |
| 13 |  MiniMax M3by MiniMax | 1306 | 55.6% | $0.309 | $1.24 | 1.0M |
| 14 | MiMo-V2.5by Xiaomi | 1303 | 55% | $0.108 | $0.288 | 1.0M |
| 15 |  GPT-5.5by OpenAI | 1296 | 56.8% | $5.15 | $30.90 | 1.1M |
| 16 | GLM 5by Z.ai | 1295 | 55.6% | $0.618 | $1.98 | 203K |
| 17 | Claude Opus 4.5by Anthropic | 1292 | 60% | $5.15 | $25.75 | 200K |
| 18 | Gemini 3.1 Pro Previewby Google | 1290 | 64.2% | $2.06 | $12.36 | 1.0M |
| 19 |  DeepSeek V4 Proby DeepSeek | 1288 | 55% | $0.448 | $0.896 | 1.0M |
| 20 | Kimi K2.5by Moonshot AI | 1286 | 54.7% | $0.386 | $2.09 | 262K |
| 21 | Claude Opus 4.8by Anthropic | 1282 | 54.9% | $5.15 | $25.75 | 1M |
| 22 | MiniMax M2.7by MiniMax | 1281 | 53.3% | $0.185 | $0.742 | 205K |
| 23 | GLM 5V Turboby Z.ai | 1278 | 52.5% | $1.24 | $4.12 | 203K |
| 24 |  Nex-N2-Proby Nex | 1274 | 50% | $0.258 | $1.03 | 262K |
| 25 | Qwen3.6 Plusby Qwen | 1273 | 51.7% | $0.335 | $2.01 | 1M |
| 26 | GLM 4.7by Z.ai | 1266 | 54.9% | $0.412 | $1.80 | 203K |
| 27 | GPT-5.4by OpenAI | 1263 | 53.4% | $2.58 | $15.45 | 1.1M |
| 28 | DeepSeek V4 Flashby DeepSeek | 1256 | 50.3% | $0.093 | $0.185 | 1.0M |
| 29 | MiniMax M2.5by MiniMax | 1256 | 56.8% | $0.124 | $0.494 | 205K |
| 30 |  Grok 4.3by xAI | 1243 | 49.6% | $1.29 | $2.58 | 1M |
Win rate is the share of head-to-head votes won on Design Arena coding tasks. Prices are per 1M tokens and include the flat 3% fee.How we rank — and what we refuse to.