Reasoning
LiveBench 2026 Data Analysis Leaderboard
LiveBench 2026 Data Analysis measures questions answered by working with tables and structured data. It belongs to the 2026 LiveBench release (first snapshot September 2, 2026), a redesigned LiveBench in which each of seven categories is scored by a single task and questions are refreshed to limit training-set contamination. It is a different test from the original LiveBench boards, so the scores are not comparable.
Source: livebench15 open models ranked+33 proprietaryData through Sep 2026
All models ranked on LiveBench 2026 Data Analysis
Proprietary / closed models are shown dimmed — you can't run them locally, but they show where the open field stands.
| # | Model | Score | VRAM @ Q4 |
|---|---|---|---|
| 1 | Gemini 4 Argon · proprietary | 85.2 | API only |
| 2 | GPT 6 Astra · proprietary | 83.0 | API only |
| 3 | GPT 6.1 Sol · proprietary | 82.7 | API only |
| 4 | GPT 5.5 · proprietary | 81.6 | API only |
| 5 | Muse Wonder 1.0 · proprietary | 81.6 | API only |
| 6 | GPT 6 Sol · proprietary | 81.2 | API only |
| 7 | Claude Fable 5 · proprietary | 80.5 | API only |
| 8 | Claude Fable 5.1 · proprietary | 80.3 | API only |
| 9 | Claude Opus 5.5 · proprietary | 80.3 | API only |
| 10 | GPT 5.6 Sol · proprietary | 79.8 | API only |
| 11 | Muse Spark 1.3 · proprietary | 79.6 | API only |
| 12 | DeepSeek V4 Flash Vision Exp · 304.6B | 79.5 | ~183 GB |
| 13 | DeepSeek V4.1 Flash · 763.2B | 79.3 | ~458 GB |
| 14 | GPT 5.6 Terra · proprietary | 79.3 | API only |
| 15 | Smaug Flash · 304.2B | 79.1 | ~609 GB |
| 16 | Kimi K3 · 2779.9B | 78.7 | ~1674 GB |
| 17 | Gemini 3.1 Pro · proprietary | 78.5 | API only |
| 18 | Qwen3.8 Max · proprietary | 78.4 | API only |
| 19 | GPT 5.6 Luna · proprietary | 78.0 | API only |
| 20 | Grok 4.7 · proprietary | 76.9 | API only |
| 21 | Qwen3.8 27B · 27.8B | 76.6 | ~17 GB |
| 22 | Mistral Large 4 · proprietary | 76.5 | API only |
| 23 | Muse Spark 1.2 · proprietary | 76.5 | API only |
| 24 | GLM 5.3 Flash · 321.3B | 76.4 | ~195 GB |
| 25 | MiniMax M3 · 427.0B | 76.2 | ~257 GB |
| 26 | Claude Sonnet 4.6 · proprietary | 76.1 | API only |
| 27 | Claude Opus 5 · proprietary | 74.6 | API only |
| 28 | Qwen3.8 Flash · proprietary | 74.2 | API only |
| 29 | Grok 4.6 · proprietary | 73.9 | API only |
| 30 | GPT 6 Luna · proprietary | 73.4 | API only |
| 31 | Grok 4.5 · proprietary | 73.0 | API only |
| 32 | Inkling · 952.4B | 72.8 | ~572 GB |
| 33 | MiMo V2.6 Pro RL · 1020B | 72.3 | ~613 GB |
| 34 | Claude Sonnet 5 · proprietary | 71.7 | API only |
| 35 | GLM 5.3 · 753.3B | 70.2 | ~456 GB |
| 36 | Gemini 3.7 Flash · proprietary | 68.0 | API only |
| 37 | Beam 1.0 · proprietary | 66.4 | API only |
| 38 | Claude Opus 4.8 · proprietary | 66.0 | API only |
| 39 | NVIDIA Nemotron 3.5 Super VL 120B A12B BF16 · 124.5B | 64.6 | ~82 GB |
| 40 | Step 5 · proprietary | 63.5 | API only |
| 41 | Muse Spark Open 816B A42B · 822.0B | 60.2 | ~543 GB |
| 42 | Claude Sonnet 5.5 · proprietary | 59.5 | API only |
| 43 | Claude Haiku 4.5 · proprietary | 59.3 | API only |
| 44 | NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B | 54.5 | ~370 GB |
| 45 | Gemini 3.8 Flash · proprietary | 54.0 | API only |
| 46 | Gemma 4 31B IT · 31.3B | 53.3 | ~20 GB |
| 47 | Claude Haiku 5.5 · proprietary | 51.8 | API only |
| 48 | NVIDIA Nemotron 3.5 Lightning 30B A3B BF16 · 31.6B | 43.1 | ~19 GB |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Qwen3.8 27B, 28B, score 76.6 — on the efficiency frontier (best score at its size or smaller).
- Smaug Flash, 304B, score 79.1 — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4 Flash Vision Exp, 305B, score 79.5 — on the efficiency frontier (best score at its size or smaller).
LiveBench 2026 Data Analysis: frequently asked questions
- What is the best open LLM on LiveBench 2026 Data Analysis?
- DeepSeek V4 Flash Vision Exp is the top open model on LiveBench 2026 Data Analysis, scoring 79.5. Among all models tested — including proprietary ones — it ranks #12. The top model overall is Gemini 4 Argon (Google) at 85.2.
- What's the best LiveBench 2026 Data Analysis model you can run on a 24 GB GPU?
- Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~17 GB), scoring 76.6 on LiveBench 2026 Data Analysis.
- Can open models match proprietary models on LiveBench 2026 Data Analysis?
- Not quite on LiveBench 2026 Data Analysis: the strongest proprietary model (Gemini 4 Argon) scores 85.2, ahead of the best open model (DeepSeek V4 Flash Vision Exp) at 79.5 — but you can run the open one yourself.
Scores aggregated from livebench. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.