Code generation and understanding
11
Models
16
Benchmarks
6
Providers
8
Open Weight
10
Deployable Now
8 self-host
11
Structured benchmark-backed rows
0
Signal-backed rows without full normalized tables
0
Benchmark-expected rows still catching up
| # | Model | Economic Footprint | Quality | Params | TerminalBench 2.0 | SWE-Bench | MMLU | Arena-Hard-Auto | $/M In | Speed | Open |
|---|---|---|---|---|---|---|---|---|---|---|---|
3 | GPT-5.1 Codex OpenAI Subscription |
| 51.3 |
| 40.6 |
| Undisclosed |
| 57.8 |
| 66.0 |
| — |
| — |
$1.25/M Cheapest verified |
| — |
4 | Qwen3 Coder 30B A3B Instruct Qwen Open WeightsStructured | 50.7 | 46.9 | 31B | — | 51.6 | 70.6 | 33.9 | $0.1500/M Cheapest verified | — |
5 | Qwen3 Next 80B A3B Instruct Qwen Open WeightsStructured | 48.2 | 45.6 | 81B | — | 42.4 | 81.9 | — | $0.0900/M Cheapest verified | — |
2 | qwen-2.5-coder-32b-instruct Qwen Open WeightsStructured | 45.8 | 35.1 | 32B | — | 40.2 | — | — | $0.6600/M Cheapest verified | — |
12 | CodeLlama 70B Meta Open WeightsStructured | 41.8 | 48.3 | 70B | — | 42.4 | — | — | Free Free | — |
7 | Qwen3 Coder Flash Qwen Open WeightsStructured | 39.3 | 20.9 | Unknown | — | 69.6 | — | — | $0.1950/M Cheapest verified | — |
6 | Qwen3 Coder Plus Qwen Open WeightsStructured | 33.3 | 20.6 | Unknown | — | — | — | — | $0.6500/M Cheapest verified | — |
14 | Grok Code Fast 1 xAI SubscriptionStructured | 30.9 | 22.1 | Undisclosed | 25.8 | — | 79.3 | — | $0.2000/M Cheapest verified | — |
10 | Devstral Medium Mistral AI Open WeightsStructured | 26.7 | 22.7 | Unknown | — | — | 70.8 | — | $0.4000/M Cheapest verified | — |
9 | North Mini Code Cohere Open WeightsStructured | 25.6 | 26.4 | 30B | 36.0 | 67.6 | — | — | Free Free | — |
11 | Codex Mini OpenAI Structured | — | — | Undisclosed | 43.1 | — | — | — | $1.50/M Cheapest verified | — |