| Metric | Value |
|---|---|
| LiteLLM wins | 277 |
| Portkey wins | 204 |
| Abstains (no tool) | 6 |
| Other tool chosen | 164 |
| Decisive cases | 481 |
| LiteLLM win rate (unweighted) | 57.6% |
| 95% CI | 53.1% - 61.9% |
| LiteLLM win rate (weighted) | 57.6% |
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 | 35 | 0 | 0 | 3% |
| DeepSeek V4 Flash | Mid | 18 | 16 | 0 | 2 | 53% |
| DeepSeek V4 Pro | Frontier | 13 | 21 | 0 | 2 | 38% |
| Llama 4 Maverick | Frontier | 26 | 5 | 1 | 4 | 84% |
| Gemini 3.5 Flash | Small | 21 | 10 | 0 | 5 | 68% |
| DeepSeek R1 0528 | Frontier | 5 | 26 | 1 | 4 | 16% |
| Claude Sonnet 4.6 | Frontier | 28 | 2 | 0 | 6 | 93% |
| MiniMax M3 | Frontier | 4 | 25 | 0 | 6 | 14% |
| MiMo V2.5 Pro | Frontier | 23 | 5 | 0 | 8 | 82% |
| Devstral 2 2512 | Mid | 12 | 15 | 0 | 1 | 44% |
| GPT 5.5 | Frontier | 21 | 4 | 0 | 8 | 84% |
| GPT 5.3 Codex | Frontier | 5 | 19 | 0 | 11 | 21% |
| GLM 5.2 | Frontier | 6 | 17 | 0 | 13 | 26% |
| GPT 5.4 Mini | Mid | 22 | 0 | 1 | 13 | 100% |
| Kimi K2.7 Code | Frontier | 21 | 1 | 0 | 3 | 95% |
| Claude Opus 4.8 | Frontier | 18 | 2 | 0 | 16 | 90% |
| Mistral Small 4 | Mid | 19 | 0 | 3 | 4 | 100% |
| Claude Haiku 4.5 | Small | 13 | 0 | 0 | 15 | 100% |
| Qwen3 Coder Next | Mid | 1 | 0 | 0 | 30 | 100% |
| Llama 4 Scout | Small | 0 | 1 | 0 | 13 | 0% |
| Prompt | Tier | LiteLLM | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-agent-application | Advanced | 50 | 43 | 0 | 13 | 54% |
| ai-engineering-workflow | Intermediate | 48 | 44 | 0 | 12 | 52% |
| ai-agent-application | Intermediate | 60 | 25 | 1 | 25 | 71% |
| ai-engineering-workflow | Advanced | 29 | 56 | 0 | 19 | 34% |
| ai-agent-application | Beginner | 71 | 12 | 1 | 30 | 86% |
| ai-engineering-workflow | Beginner | 19 | 24 | 4 | 65 | 44% |
