Best AI with web search: 42 models ranked by blind votes
Models with live web search answer the same question, and voters pick the answer they prefer. Ranked by LMArena’s blind votes, every score with its margin of error and, beside it, the list price of each model on our price list.
Scores from LMArena, published 13 October 2026, under CC BY 4.0. Prices from each maker’s pricing page, read on 15 October 2026.
- Models ranked
- 42
- Level with the leader
- 3
- On our price list
- 10
- Published
- 13 October 2026
Level: the votes cannot separate it from the leader.
On LMArena’s Search board, GPT-5.6 Sol from OpenAI leads with 1258 points from 60,437 blind votes, published 13 October 2026. 2 other models are statistically level with it. It also tops the models with an API price, at $8 per million tokens with input and output weighed 3 to 1.
Best AI with web search
Models with live web search answer the same question, and voters pick the answer they prefer.
| # | Model | Score | Votes | Per 1M tokens, in / out |
|---|---|---|---|---|
| 1 | GPT-5.6 SolOpenAI · gpt-5.6-sol-xhigh | 95% interval 1252 to 1264 | 60,437 | $4.00 / $20.00 |
| 1 | GPT-6.1 SolOpenAI · gpt-6.1-sol-max | 95% interval 1241 to 1259 | 5,555 | $2.00 / $10.00 |
| 1 | GPT-6 AstraOpenAI · gpt-6-astra-max | 95% interval 1238 to 1254 | 18,483 | $10.00 / $50.00 |
| 2 | Qwen3.8 MaxAlibaba · qwen3.8-max | 95% interval 1216 to 1243 | 1,870 | $2.00 / $6.00 |
| 3 | Claude Fable 5.1Anthropic · claude-fable-5.1-high | 95% interval 1226 to 1241 | 12,782 | $10.00 / $50.00 |
| 4 | claude-fable-5-highAnthropic | 95% interval 1222 to 1236 | 73,596 | · |
| 4 | GPT-6 SolOpenAI · gpt-6-sol-max | 95% interval 1221 to 1236 | 11,313 | $2.00 / $10.00 |
| 4 | gpt-5.5-searchOpenAI | 95% interval 1222 to 1231 | 125,957 | · |
| 4 | claude-opus-4-6-searchAnthropic | 95% interval 1217 to 1227 | 173,191 | · |
| 7 | Claude Opus 5Anthropic · claude-opus-5-high | 95% interval 1208 to 1221 | 13,285 | $5.00 / $25.00 |
| 8 | Claude Opus 5.5Anthropic · claude-opus-5.5-medium | 95% interval 1204 to 1218 | 10,842 | $4.00 / $20.00 |
| 9 | GPT-6 LunaOpenAI · gpt-6-luna-max | 95% interval 1202 to 1217 | 10,793 | $0.10 / $0.50 |
| 10 | claude-opus-4-7Anthropic | 95% interval 1205 to 1215 | 127,506 | · |
| 10 | gemini-3.1-pro-groundingGoogle | 95% interval 1202 to 1211 | 151,436 | · |
| 10 | grok-4.20-multi-agent-beta-0309xAI | 95% interval 1201 to 1212 | 126,238 | · |
| 10 | grok-4.5xAI | 95% interval 1200 to 1212 | 76,458 | · |
| 11 | claude-sonnet-4-6-searchAnthropic | 95% interval 1197 to 1207 | 173,986 | · |
| 11 | gemini-3-pro-groundingGoogle | 95% interval 1195 to 1206 | 38,857 | · |
| 11 | grok-4.20-beta1xAI | 95% interval 1193 to 1205 | 57,049 | · |
| 13 | gpt-5.4-searchOpenAI | 95% interval 1191 to 1202 | 126,730 | · |
| 13 | o3-searchOpenAI | 95% interval 1189 to 1203 | 20,770 | · |
| 15 | claude-opus-4-8Anthropic | 95% interval 1190 to 1201 | 106,304 | · |
| 15 | ernie-5.1Baidu | 95% interval 1183 to 1202 | 4,024 | · |
| 17 | grok-4-1-fast-searchxAI | 95% interval 1190 to 1200 | 85,919 | · |
| 17 | claude-sonnet-5-searchAnthropic | 95% interval 1187 to 1199 | 72,978 | · |
| 17 | Grok 4.3xAI · grok-4.3 | 95% interval 1188 to 1198 | 128,393 | $1.25 / $2.50 |
| 19 | gpt-5-searchOpenAI | 95% interval 1180 to 1195 | 20,865 | · |
| 20 | gemini-3-flash-groundingGoogle | 95% interval 1184 to 1193 | 196,399 | · |
| 20 | claude-opus-4-5-searchAnthropic | 95% interval 1181 to 1192 | 65,047 | · |
| 24 | gpt-5.1-searchOpenAI | 95% interval 1178 to 1189 | 63,042 | · |
| 28 | claude-sonnet-4-5-searchAnthropic | 95% interval 1171 to 1181 | 144,153 | · |
| 31 | claude-opus-4-1-searchAnthropic | 95% interval 1162 to 1172 | 79,806 | · |
| 32 | gpt-5.2-searchOpenAI | 95% interval 1156 to 1167 | 55,634 | · |
| 32 | grok-4-fast-searchxAI | 95% interval 1156 to 1166 | 43,083 | · |
| 33 | claude-opus-4-searchAnthropic | 95% interval 1146 to 1157 | 31,506 | · |
| 35 | gemini-2.5-pro-groundingGoogle | 95% interval 1139 to 1149 | 86,689 | · |
| 35 | ppl-sonar-pro-highPerplexity | 95% interval 1137 to 1149 | 28,899 | · |
| 36 | ppl-sonar-reasoning-pro-highPerplexity | 95% interval 1133 to 1144 | 29,439 | · |
| 36 | gpt-5.2-search-non-reasoningOpenAI | 95% interval 1130 to 1142 | 79,796 | · |
| 40 | grok-4-searchxAI | 95% interval 1117 to 1129 | 19,195 | · |
| 41 | api-gpt-4o-searchOpenAI | 95% interval 1072 to 1095 | 3,399 | · |
| 42 | diffbot-small-xlDiffbot · open weights | 95% interval 1054 to 1071 | 6,368 | · |
Worth paying for: the models no cheaper one beats
The models with an API price that no cheaper model outscores on this board, cheapest first. Every other priced model costs more than one of these and scores no higher. The price weighs input and output 3 to 1, the usual mix of a chat or document task.
| Model | Score | Rank | Per 1M tokens, 3:1 |
|---|---|---|---|
| GPT-6 LunaOpenAI · gpt-6-luna-max | 1209 | 9 of 42 | |
| Qwen3.8 MaxAlibaba · qwen3.8-max | 1230 | 2 of 42 | |
| GPT-6.1 SolOpenAI · gpt-6.1-sol-max | 1250 | 1 of 42 | |
| GPT-5.6 SolOpenAI · gpt-5.6-sol-xhigh | 1258 | 1 of 42 |
Where the ranking comes from
The scores, intervals and ranks are LMArena’s. The prices are ours.
- Blind votes
- LMArena puts the same prompt to two anonymous models and asks the person who wrote it to pick the better answer. The votes place every model on one scale, where a gap of 100 points means the higher model wins about 64 times in 100.
- Style control
- The scores discount answer length and formatting, so a model does not climb by writing more or by adding headings and lists.
- Shared ranks
- A model’s rank is one plus the number of models whose whole 95% interval sits above its own. Models the votes cannot separate share a rank, which is why several can be first.
- Settings
- A model can appear more than once, at different reasoning settings: the line under its name gives the entry as LMArena lists it. Entries with web search switched on are a different product and carry no price.
- Prices
- The list price per million tokens from the maker’s own pricing page, read on 15 October 2026. It is a price per token, not per answer: a higher reasoning setting spends more tokens on the same question.
Source: LMArena, leaderboard dataset (revision 4589dee8, published 13 October 2026), licensed under CC BY 4.0. We change no score: we add the prices and the links to our model pages.
What each of them costs to run
The price list works three everyday jobs on every model, and our own benchmark runs them on real business tasks to see which pass.
Scores published 13 October 2026