Reported by others
Answering by searching the web (DeepResearch Bench)
Research tasks answered by searching the web, such as finding a number, checking a claim or collecting data.
How to read this table
- Higher rows are rated higher. A range like “1–2” means the gap is within the margin of error.
- The score is the value published by the task's authors and other organizations.
- Conditions can differ between models, so small gaps do not mean much.
- Prices are a guide per million tokens, reading then writing.
- Length is how much text the model can read at once (tokens).
| Rank | Model | Developer | Score | Input $ /1M | Output $ /1M | Context | Supports | Weights | Released | Price and specs |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 | Anthropic | Rating 55.3 | $5 | $25 | 1M | Anthropic Released 2026-02 Price $5 / $25 Length 1M | |||
| 2 | Claude Sonnet 4.6 | Anthropic | Rating 54.9 | $3 | $15 | 1M | Anthropic Released 2026-02 Price $3 / $15 Length 1M | |||
| 3 | Claude Opus 4.5 | Anthropic | Rating 54.8 | $5 | $25 | 200K | Anthropic Released 2025-11 Price $5 / $25 Length 200K | |||
| 4 | GPT-5.5 | OpenAI | Rating 54.0 | $5 | $30 | 1.05M | OpenAI Released 2026-04 Price $5 / $30 Length 1.05M | |||
| 5 | Claude Sonnet 4.5 | Anthropic | Rating 52.6 | $3 | $15 | 200K | Anthropic Released 2025-09 Price $3 / $15 Length 200K | |||
| 6 | Claude Opus 4.8 | Anthropic | Rating 50.2 | $5 | $25 | 1M | Anthropic Released 2026-05 Price $5 / $25 Length 1M | |||
| 7 | Gemini 3 Flash (preview) | Rating 49.8 | $0.5 | $3 | 1M | Google Released 2025-12 Price $0.5 / $3 Length 1M | ||||
| 8 | GPT-5 | OpenAI | Rating 49.6 | $1.25 | $10 | 400K | OpenAI Released 2025-08 Price $1.25 / $10 Length 400K | |||
| 9 | Claude Opus 4.1 | Anthropic | Rating 48.3 | — | — | 200K | Anthropic Released 2025-08 Price — Length 200K | |||
| 10 | Gemini 3.1 Pro (preview) | Rating 47.8 | $2 | $12 | 1.05M | Google Released 2026-02 Price $2 / $12 Length 1.05M | ||||
| 11 | Grok 4 | xAI | Rating 47.3 | — | — | 256K | xAI Released 2025-07 Price — Length 256K | |||
| 12 | Claude Opus 4 | Anthropic | Rating 46.8 | — | — | 200K | Anthropic Released 2025-05 Price — Length 200K | |||
| 13 | Claude Sonnet 4 | Anthropic | Rating 46.6 | — | — | 200K | Anthropic Released 2025-05 Price — Length 200K | |||
| 14 | Gemini 3 Pro (preview) | Rating 46.3 | — | — | 1M | Google Released 2025-11 Price — Length 1M | ||||
| 15 | Claude Haiku 4.5 | Anthropic | Rating 45.5 | $1 | $5 | 200K | Anthropic Released 2025-10 Price $1 / $5 Length 200K | |||
| 16 | o3 | OpenAI | Rating 45.2 | $2 | $8 | 200K | OpenAI Released 2025-04 Price $2 / $8 Length 200K | |||
| 17 | Claude 3.7 Sonnet | Anthropic | Rating 43.6 | — | — | 200K | Anthropic Released 2025-02 Price — Length 200K | |||
| 18–19 | Gemini 2.5 Pro (preview, 2025-05) | Rating 42.8 | — | — | 1.05M | Google Released 2025-05 Price — Length 1.05M | ||||
| 18–19 | GPT-5.1 | OpenAI | Rating 42.8 | $1.25 | $10 | 400K | OpenAI Released 2025-11 Price $1.25 / $10 Length 400K | |||
| 20 | Gemini 2.5 Pro | Rating 41.5 | $1.25 | $10 | 1M | Google Released 2025-06 Price $1.25 / $10 Length 1M | ||||
| 21 | GPT-5.2 | OpenAI | Rating 41.1 | $1.75 | $14 | 400K | OpenAI Released 2025-12 Price $1.75 / $14 Length 400K | |||
| 22 | Gemini 3.1 Flash-Lite | Rating 37.3 | $0.25 | $1.5 | 1.05M | Google Released 2026-03 Price $0.25 / $1.5 Length 1.05M | ||||
| 23 | GPT-5.4 mini | OpenAI | Rating 36.3 | $0.75 | $4.5 | 400K | OpenAI Released 2026-03 Price $0.75 / $4.5 Length 400K | |||
| 24–25 | GPT-5.4 | OpenAI | Rating 35.1 | $2.5 | $15 | 1.05M | OpenAI Released 2026-03 Price $2.5 / $15 Length 1.05M | |||
| 24–25 | DeepSeek-R1 (2025-05) | DeepSeek | Rating 35.1 | — | — | 131.1K | DeepSeek Released 2025-05 Price — Length 131.1K |
Best score over time
2025-02-24Release date2026-05-28
How to read this page
- Each answer is graded for correctness, and the value is the average score. Figures published by the organizations that made the tasks, among others, collected here.
- The average correctness grade per task (out of 100), and most models score 35 to 55. No error margins are published, so differences of 1 to 2 points are not clear differences.
- In the Supports column the icons are, from left: can use tools / can read images / can answer in a fixed format (JSON); faded means not supported or unknown.
- Open-weights models have their contents (the weights) published, so you can also run them on your own computer or server. Models without the open-weights mark are normally used through the developer's own services (apps or API); for some models the weights status is unknown.
- Names such as ChatGPT next to a model are the consumer services its developer offers. Check on the service itself whether that model is available there.
Data: Arena, Epoch AI, Swallow LLM team (CC BY 4.0), Models.dev (MIT), updated 2026-10-24, data sources