Best LLM
What is the best LLM for research?
In LMArena Search, GPT-5.6 Sol is ahead with a rating of 1,261.
The ranking rests on this one measurement alone, a rating from publicly collected pairwise votes, and is not an area of the Gradually overall index.
Among the ten best, Grok 4.5 is the cheapest at $3 per million tokens.
The fastest among them is Gemini 3.1 Pro Preview, at 96 tokens per second.
- Measured models
- 14
- Data as of
- Oct 30, 2026
Ranking: Research
The ranking rests on LMArena Search alone, a rating from publicly collected pairwise votes run by LMArena. The data contains no dedicated test for research, more in the methodology.
| Rank | Model | LMArena Search | Overall index | Price (3:1) | Speed |
|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol | 1,261 | 109.1 | $8 | 75 |
| 2 | GPT-5.5 | 1,247 | 101.5 | $11.25 | 87 |
| 3 | Claude Opus 5 | 1,242 | 117.4 | $10 | 71 |
| 4 | GPT-6 Astra | 1,234 | 120.2 | $20 | 58 |
| 5 | GPT-6.1 Sol | 1,233 | 118.0 | $4 | 44 |
| 6 | Claude Opus 4.7 | 1,230 | 95.1 | $10 | 57 |
| 7 | Claude Fable 5 | 1,229 | 112.8 | $20 | 50 |
| 8 | Grok 4.5 | 1,216 | 94.8 | $3 | 52 |
| 9 | Claude Opus 5.5 | 1,213 | 125.7 | $8 | 75 |
| 10 | Gemini 3.1 Pro Preview | 1,209 | 91.4 | $4.50 | 96 |
| 11 | Claude Opus 4.8 | 1,204 | 99.5 | $10 | 71 |
| 12 | GPT-6 Sol | 1,175 | 109.1 | $4 | 82 |
| 13 | Grok 4.3 | 1,166 | 71.6 | $1.56 | 66 |
| 14 | GPT-6 Luna | 1,160 | 95.8 | $0.20 | 77 |
The 14 are the models with a value in LMArena Search. The overall index counts different sets of models, so the numbers differ between pages.
Price in dollars per million tokens, mixed from 3 parts input and 1 part output. Speed in tokens per second.
Which tests count here?
LMArena Search
14 models tested · LMArena