LMArena Vision: homework leaderboard
- models measured
- 74
- lead over #2
- +1 Elo
- median
- 1286
Leader on LMArena Vision: homework · read October 21
On October 21, 2026, 74 of our catalog's models had a LMArena Vision: homework result, and Anthropic Claude Opus 5.5 led with 1362.
Image prompts with homework problems. People ask a question, compare two anonymous answers and vote for the better one; the rating is built from those votes, as LMArena publishes it.
Role in the index
We collect this benchmark and show it as supporting evidence. It is not part of the Vibe Coding Index, because it measures something other than coding and the index is about coding. Read the methodology
| # | Model | Result | Votes | Interval (±) | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 1362 | 422 | ±29.4 | 92.2 | $20.00 | |
| 2 | Claude Fable 5Anthropic | 1361 | 1,751 | ±14.7 | 82.1 | $50.00 | |
| 3 | GPT-6.1 SolOpenAI | 1356 | 368 | ±31.0 | 88.7 | $10.00 | |
| 4 | Gemini 3.8 FlashGoogle | 1351 | 681 | ±22.4 | 70.5 | $4.125 | |
| 5 | Claude Fable 5.1Anthropic | 1345 | 756 | ±21.4 | 89.1 | $50.00 | |
| 6 | Claude Opus 5Anthropic | 1338 | 1,899 | ±14.0 | 81.8 | $25.00 | |
| 7 | GPT-6 AstraOpenAI | 1337 | 560 | ±24.8 | 88.7 | $50.00 | |
| 8 | Claude Opus 4.7Anthropic | 1337 | 3,084 | ±11.6 | 51.1 | $25.00 | |
| 9 | Claude Opus 4.8Anthropic | 1337 | 2,077 | ±13.5 | 66.0 | $25.00 | |
| 10 | Claude Opus 4.6Anthropic | 1337 | 2,845 | ±11.9 | 66.9 | $25.00 | |
| 11 | GPT-5.4OpenAI | 1336 | 4,003 | ±10.3 | 37.7 | $15.00 | |
| 12 | Qwen3.8 Max (0902)Qwen | 1336 | 909 | ±19.4 | 63.5 | $6.00 | |
| 13 | GPT-5.5OpenAI | 1332 | 3,535 | ±10.8 | 67.8 | $30.00 | |
| 14 | Claude Sonnet 5.5Anthropic | 1331 | 395 | ±29.6 | 88.9 | $10.00 | |
| 15 | Gemini 3.5 FlashGoogle | 1330 | 1,553 | ±15.3 | 42.8 | $9.00 | |
| 16 | Gemini 3.7 FlashGoogle | 1320 | 665 | ±22.7 | 69.9 | $3.75 | |
| 17 | GLM 5.3 FlashZ.ai | 1318 | 2,229 | ±12.8 | 71.0 | $0.50 | |
| 18 | Gemini 3.1 Pro PreviewGoogle | 1316 | 6,293 | ±8.4 | 35.3 | $12.00 | |
| 19 | Gemini 3.6 FlashGoogle | 1314 | 993 | ±18.6 | 48.3 | $3.75 | |
| 20 | GPT-5.6 TerraOpenAI | 1312 | 1,286 | ±16.7 | 69.0 | $12.00 | |
| 21 | Claude Sonnet 4.6Anthropic | 1311 | 3,700 | ±10.7 | 57.8 | $15.00 | |
| 22 | Claude Sonnet 5Anthropic | 1311 | 1,500 | ±15.7 | 63.5 | $10.00 | |
| 23 | Kimi K2.6Moonshot AI | 1310 | 2,263 | ±13.0 | 44.2 | $3.41 | |
| 24 | Gemma 4 31BGoogle | 1310 | 6,435 | ±8.7 | 38.3 | $0.40 | |
| 25 | Grok 4.7xAI | 1309 | 517 | ±25.4 | 75.0 | $6.00 | |
| 26 | DeepSeek V4.1 FlashDeepSeek | 1307 | 521 | ±25.5 | 77.2 | $0.525 | |
| 27 | GPT-5.6 SolOpenAI | 1305 | 1,190 | ±17.3 | 76.7 | $20.00 | |
| 28 | Kimi K2.5Moonshot AI | 1297 | 3,649 | ±10.2 | 58.5 | $2.50 | |
| 29 | Muse Spark 1.3Meta | 1296 | 680 | ±22.9 | 79.7 | $4.25 | |
| 30 | Grok 4.5xAI | 1295 | 1,284 | ±17.1 | 75.5 | $6.00 | |
| 31 | Qwen3.7 PlusQwen | 1295 | 1,646 | ±14.8 | Not benchmarked | $1.28 | |
| 32 | Qwen3.8 27BQwen | 1295 | 919 | ±19.8 | 73.6 | $1.35 | |
| 33 | Gemma 4 26B A4BGoogle | 1293 | 3,822 | ±10.4 | 22.8 | $0.255 | |
| 34 | Qwen3.5 397B A17BQwen | 1292 | 4,001 | ±9.7 | 28.6 | $2.34 | |
| 35 | Gemini 3.5 Flash LiteGoogle | 1291 | 1,010 | ±19.0 | 33.6 | $2.50 | |
| 36 | Muse Spark 1.1Meta | 1287 | 1,354 | ±16.2 | 48.9 | $4.25 | |
| 37 | GPT-6 SolOpenAI | 1286 | 518 | ±25.8 | 82.8 | $10.00 | |
| 38 | GPT-5.4 MiniOpenAI | 1285 | 3,492 | ±10.9 | 28.2 | $4.50 | |
| 39 | Grok 4.6xAI | 1283 | 873 | ±19.9 | 76.4 | $6.00 | |
| 40 | MiMo-V2.6-FlashXiaomi | 1283 | 546 | ±25.6 | 61.9 | $0.28 | |
| 41 | GPT-5.6 LunaOpenAI | 1282 | 1,128 | ±18.2 | 61.5 | $1.20 | |
| 42 | Gemini 2.5 ProGoogle | 1281 | 5,415 | ±8.8 | 37.0 | $10.00 | |
| 43 | GPT-5OpenAI | 1278 | 1,325 | ±16.8 | 72.2 | $10.00 | |
| 44 | Qwen3 VL 235B A22B InstructQwen | 1277 | 1,611 | ±14.5 | Not benchmarked | $1.90 | |
| 45 | GPT-5.2OpenAI | 1276 | 2,888 | ±11.6 | 58.9 | $14.00 | |
| 46 | GLM 5V TurboZ.ai | 1273 | 5,226 | ±9.3 | 28.8 | $4.00 | |
| 47 | GPT-5.1OpenAI | 1272 | 1,218 | ±16.6 | 53.3 | $10.00 | |
| 48 | Gemini 3.1 Flash Lite PreviewGoogle | 1271 | 3,947 | ±10.1 | 8.1 | $1.50 | |
| 49 | MiniMax M3MiniMax | 1268 | 2,405 | ±12.7 | 57.8 | $1.20 | |
| 50 | Step 5 PreviewStepFun | 1265 | 233 | ±37.8 | 71.4 | $2.70 | |
| 51 | MiMo-V2.6-ProXiaomi | 1264 | 226 | ±39.4 | 72.6 | $0.87 | |
| 52 | MiMo-V2.5Xiaomi | 1262 | 4,076 | ±10.1 | 34.0 | $0.238 | |
| 53 | Qwen3.5-122B-A10BQwen | 1261 | 1,929 | ±13.8 | 22.9 | $2.08 | |
| 54 | GPT-6 LunaOpenAI | 1261 | 550 | ±25.3 | 71.4 | $0.50 | |
| 55 | T Inkling SmallThinking Machines | 1257 | 835 | ±20.4 | 29.9 | $1.20 | |
| 56 | Qwen3.5-27BQwen | 1256 | 3,843 | ±10.0 | 22.6 | $2.16 | |
| 57 | GPT-4.1OpenAI | 1248 | 1,478 | ±16.4 | 46.5 | $8.00 | |
| 58 | Gemini 2.5 FlashGoogle | 1244 | 4,863 | ±9.0 | 28.7 | $2.50 | |
| 59 | Grok 4.3xAI | 1241 | 3,560 | ±10.6 | 22.5 | $2.50 | |
| 60 | o4 MiniOpenAI | 1234 | 1,588 | ±16.0 | 68.6 | $4.40 | |
| 61 | o3OpenAI | 1234 | 1,930 | ±14.8 | 70.7 | $8.00 | |
| 62 | Mistral Medium 3.5Mistral AI | 1230 | 803 | ±22.0 | 9.2 | $7.50 | |
| 63 | GPT-5.4 NanoOpenAI | 1225 | 3,429 | ±11.2 | Not benchmarked | $1.25 | |
| 64 | GPT-4.1 MiniOpenAI | 1214 | 1,310 | ±17.2 | 28.5 | $1.60 | |
| 65 | GPT-5 MiniOpenAI | 1210 | 976 | ±19.4 | 59.8 | $2.00 | |
| 66 | Mistral Medium 3.1Mistral AI | 1165 | 2,146 | ±13.4 | Not benchmarked | $2.00 | |
| 67 | Mistral Medium 3Mistral AI | 1165 | 809 | ±20.7 | Not benchmarked | $2.00 | |
| 68 | Llama 4 MaverickMeta Llama | 1164 | 526 | ±25.2 | 18.1 | $0.6525 | |
| 69 | GLM 4.5VZ.ai | 1163 | 215 | ±38.4 | Not benchmarked | $1.80 | |
| 70 | Gemma 3 27BGoogle | 1160 | 1,034 | ±18.8 | 4.9 | $0.35 | |
| 71 | Mistral Small 3.1 24BMistral AI | 1158 | 809 | ±21.1 | Not benchmarked | $0.555 | |
| 72 | GPT-5 NanoOpenAI | 1157 | 260 | ±34.9 | 34.8 | $0.40 | |
| 73 | GLM 4.6VZ.ai | 1155 | 271 | ±36.4 | Not benchmarked | $0.90 | |
| 74 | Llama 4 ScoutMeta Llama | 1148 | 477 | ±26.0 | 9.1 | $0.30 |
Only models in our catalog are listed, and each appears once with its best published run. 172 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.