| Metric | Value |
|---|---|
| LiteLLM wins | 328 |
| Portkey wins | 228 |
| Abstains (no tool) | 7 |
| Other tool chosen | 189 |
| Decisive cases | 556 |
| LiteLLM win rate (unweighted) | 59.0% |
| 95% CI | 54.9% - 63.0% |
| LiteLLM win rate (weighted) | 59.0% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | LiteLLM | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| Gemini 2.5 Pro | Frontier | 1 | 41 | 0 | 0 | 2% |
| DeepSeek V4 Flash | Mid | 20 | 19 | 0 | 3 | 51% |
| DeepSeek V4 Pro | Frontier | 16 | 23 | 0 | 3 | 41% |
| Llama 4 Maverick | Frontier | 32 | 5 | 1 | 4 | 86% |
| Gemini 3.5 Flash | Small | 25 | 11 | 0 | 6 | 69% |
| DeepSeek R1 0528 | Frontier | 6 | 30 | 1 | 5 | 17% |
| Claude Sonnet 4.6 | Frontier | 33 | 2 | 0 | 7 | 94% |
| MiniMax M3 | Frontier | 5 | 28 | 0 | 7 | 15% |
| MiMo V2.5 Pro | Frontier | 27 | 5 | 0 | 9 | 84% |
| GPT 5.4 Mini | Mid | 28 | 0 | 1 | 13 | 100% |
| GPT 5.5 | Frontier | 24 | 4 | 0 | 10 | 86% |
| GLM 5.2 | Frontier | 9 | 19 | 0 | 14 | 32% |
| GPT 5.3 Codex | Frontier | 6 | 22 | 0 | 13 | 21% |
| Kimi K2.7 Code | Frontier | 26 | 1 | 0 | 3 | 96% |
| Devstral 2 2512 | Mid | 12 | 15 | 0 | 1 | 44% |
| Claude Opus 4.8 | Frontier | 22 | 2 | 0 | 18 | 92% |
| Mistral Small 4 | Mid | 21 | 0 | 4 | 5 | 100% |
| Claude Haiku 4.5 | Small | 14 | 0 | 0 | 17 | 100% |
| Qwen3 Coder Next | Mid | 1 | 0 | 0 | 36 | 100% |
| Llama 4 Scout | Small | 0 | 1 | 0 | 15 | 0% |
| Prompt | Tier | LiteLLM | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-agent-application | Advanced | 61 | 48 | 0 | 16 | 56% |
| ai-engineering-workflow | Intermediate | 58 | 49 | 0 | 15 | 54% |
| ai-agent-application | Intermediate | 73 | 27 | 1 | 28 | 73% |
| ai-engineering-workflow | Advanced | 34 | 62 | 0 | 21 | 35% |
| ai-agent-application | Beginner | 80 | 15 | 1 | 34 | 84% |
| ai-engineering-workflow | Beginner | 22 | 27 | 5 | 75 | 45% |
