AI BENCHY Category
Data parsing and extraction Ranking
See which AI models perform best on Data parsing and extraction, which ones stay reliable, and where the biggest gaps appear.
Related Failure Reasons
| Rank | Model | Company | Data parsing and extraction Score | Avg Score | Tests Correct | Response Time (avg) |
|---|---|---|---|---|---|---|
| #13 | Step 3.5 Flash medium | Stepfun | 10.0 | 7.4 | 2/2 | 15.0s |
| #34 | GPT-5 Nano medium | OpenAI | 10.0 | 5.5 | 0/2 | 21.4s |
| #43 | MiniMax M2.5 medium | Minimax | 10.0 | 4.7 | 0/2 | 7.48s |
| #54 | MiMo-V2-Flash none | Xiaomi | 10.0 | 2.9 | 0/2 | 19.7s |
| #55 | LFM2-24B-A2B none | Liquid | 10.0 | 2.6 | 0/2 | 714ms |
| #1 | Gemini 3 Flash Preview medium | 9.9 | 10.0 | 2/2 | 4.72s | |
| #2 | Gemini 3.1 Pro Preview medium | 9.9 | 9.4 | 2/2 | 7.72s | |
| #3 | GPT-5.3-Codex medium | OpenAI | 9.9 | 8.4 | 2/2 | 3.07s |
| #4 | Qwen3.5 Plus 2026-02-15 medium | Qwen | 9.9 | 8.3 | 2/2 | 46.9s |
| #5 | Gemini 3 Flash Preview low | 9.9 | 8.2 | 2/2 | 9.40s | |
| #6 | Gemini 3 Pro Preview medium | 9.9 | 8.2 | 2/2 | 10.8s | |
| #7 | Qwen3.5-27B medium | Qwen | 9.9 | 8.2 | 2/2 | 30.3s |
| #8 | Gemini 3.1 Flash Lite Preview high | 9.9 | 8.2 | 2/2 | 7.16s | |
| #9 | GPT-5.4 medium | OpenAI | 9.9 | 8.0 | 2/2 | 5.32s |
| #10 | Qwen3.5-122B-A10B medium | Qwen | 9.9 | 7.7 | 2/2 | 23.4s |
| #11 | Claude Sonnet 4.6 medium | Anthropic | 9.9 | 7.7 | 2/2 | 13.9s |
| #12 | Gemini 3.1 Flash Lite Preview medium | 9.9 | 7.5 | 2/2 | 2.29s | |
| #15 | GPT-5.2 Chat none | OpenAI | 9.9 | 7.4 | 2/2 | 3.05s |
| #16 | Gemini 2.5 Flash medium | 9.9 | 7.4 | 2/2 | 4.06s | |
| #17 | Gemini 3.1 Flash Lite Preview low | 9.9 | 7.3 | 2/2 | 3.00s | |
| #18 | DeepSeek V3.2 medium | DeepSeek | 9.9 | 7.3 | 2/2 | 36.1s |
| #19 | GPT-5.3 Chat none | OpenAI | 9.9 | 7.3 | 2/2 | 2.21s |
| #20 | Gemini 3 Flash Preview none | 9.9 | 7.2 | 2/2 | 1.41s | |
| #22 | Gemini 3.1 Flash Lite Preview none | 9.9 | 7.1 | 2/2 | 1.22s | |
| #23 | Seed-2.0-Mini medium | Bytedance Seed | 9.9 | 6.9 | 2/2 | 24.3s |
| #25 | Claude Sonnet 4.6 none | Anthropic | 9.9 | 6.8 | 2/2 | 3.43s |
| #26 | Claude Opus 4.6 medium | Anthropic | 9.9 | 6.6 | 2/2 | 7.37s |
| #27 | GPT-5.2 medium | OpenAI | 9.9 | 6.5 | 2/2 | 3.15s |
| #28 | Kimi K2.5 medium | Moonshot AI | 9.9 | 6.4 | 2/2 | 49.8s |
| #29 | Qwen3.5 Plus 2026-02-15 none | Qwen | 9.9 | 6.2 | 2/2 | 1.89s |
| #30 | Grok 4.1 Fast medium | X AI | 9.9 | 6.2 | 2/2 | 6.63s |
| #31 | GLM 5 none | Z.ai | 9.9 | 6.0 | 2/2 | 5.78s |
| #32 | GPT-5 Mini medium | OpenAI | 9.9 | 6.0 | 2/2 | 12.6s |
| #37 | Qwen3.5-Flash none | Qwen | 9.9 | 5.2 | 2/2 | 1.57s |
| #38 | Gemini 2.5 Flash none | 9.9 | 5.2 | 2/2 | 652ms | |
| #40 | Qwen3.5-122B-A10B none | Qwen | 9.9 | 5.0 | 2/2 | 1.01s |
| #41 | Qwen3.5-27B none | Qwen | 9.9 | 4.9 | 2/2 | 1.43s |
| #42 | Qwen3.5-35B-A3B none | Qwen | 9.9 | 4.7 | 2/2 | 1.16s |
| #44 | GPT-5.4 none | OpenAI | 9.9 | 4.5 | 2/2 | 1.04s |
| #45 | Trinity Large Preview none | Arcee AI | 9.9 | 4.2 | 2/2 | 3.26s |
| #47 | GPT-4o-mini none | OpenAI | 9.9 | 4.0 | 2/2 | 1.27s |
| #53 | Grok 4.1 Fast none | X AI | 9.9 | 2.9 | 2/2 | 943ms |
| #21 | MiMo-V2-Flash medium | Xiaomi | 5.5 | 7.2 | 1/2 | 0ms |
| #24 | Qwen3.5-Flash medium | Qwen | 5.5 | 6.9 | 1/2 | 57.0s |
| #35 | Qwen3.5-35B-A3B medium | Qwen | 5.5 | 5.5 | 1/2 | 59.3s |
| #36 | Mercury 2 medium | Inception | 5.5 | 5.3 | 1/2 | 1.11s |
| #39 | gpt-oss-120b medium | OpenAI | 5.5 | 5.1 | 1/2 | 1.98s |
| #51 | Mercury 2 none | Inception | 5.5 | 3.4 | 1/2 | 667ms |
| #33 | DeepSeek V3.2 none | DeepSeek | 5.4 | 5.5 | 1/2 | 9.42s |
| #46 | Kimi K2.5 none | Moonshot AI | 5.4 | 4.1 | 1/2 | 42.1s |
| #48 | Qwen3 Coder Next none | Qwen | 5.4 | 4.0 | 1/2 | 1.32s |
| #49 | GLM 4.7 Flash none | Z.ai | 5.4 | 3.9 | 1/2 | 4.82s |
| #50 | Qwen3 Coder Next medium | Qwen | 5.4 | 3.5 | 1/2 | 81.8s |
| #14 | GLM 5 medium | Z.ai | 5.0 | 7.4 | 1/2 | 8.90s |
| #52 | GLM 4.7 Flash medium | Z.ai | 5.0 | 3.1 | 1/2 | 1.51s |