Skip to main content

Human ratings

Writing and language work (human ratings)

Among the occupational fields questions are sorted into, a category computed only from questions in writing and language work.

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 rating comes from people comparing two answers and picking the better one.
  • At a gap of 50, the higher model is picked about 57% of the time; 100 means 64%, 200 means 76%.
  • Prices are a guide per million tokens, reading then writing.
  • Length is how much text the model can read at once (tokens).

± range of the top model

RankModelDeveloperratingrating scale 1400 to 1550Week
change
OverallInput $
/1M
Output $
/1M
ContextSupportsWeightsReleasedPrice and specs
1–4Claude Opus 5.5AnthropicRating 1520 ±1201509rank 1–7$4$201MAnthropic Released 2026-⁠09 Price $4 / $20 Length 1M
1–6Gemini 4 ArgonGoogleRating 1515 ±12−61517rank 1–3$2$101.05MGoogle Released 2026-⁠09 Week change −6 Price $2 / $10 Length 1.05M
1–7Claude Fable 5AnthropicRating 1508 ±601503rank 2–9$10$501MAnthropic Released 2026-⁠06 Price $10 / $50 Length 1M
1–14Muse WonderMetaRating 1501 ±15+61509rank 1–8$4$201.05MMeta Released 2026-⁠10 Week change +6 Price $4 / $20 Length 1.05M
2–9Claude Opus 4.6AnthropicRating 1499 ±5−11503rank 2–9$5$251MAnthropic Released 2026-⁠02 Week change −1 Price $5 / $25 Length 1M
3–9Claude Opus 4.7AnthropicRating 1497 ±601501rank 2–10$5$251MAnthropic Released 2026-⁠04 Price $5 / $25 Length 1M
2–14Claude Fable 5.1AnthropicRating 1495 ±10+41503rank 2–10$10$501MAnthropic Released 2026-⁠09 Week change +4 Price $10 / $50 Length 1M
4–14Gemini 3.7 FlashGoogleRating 1493 ±801490rank 8–25$0.75$3.751.05MGoogle Released 2026-⁠08 Price $0.75 / $3.75 Length 1.05M
4–14Gemini 3.8 FlashGoogleRating 1491 ±7−11496rank 4–15$0.75$3.751.05MGoogle Released 2026-⁠09 Week change −1 Price $0.75 / $3.75 Length 1.05M
6–21Gemini 3.1 Pro (preview)GoogleRating 1482 ±501487rank 10–26$2$121.05MGoogle Released 2026-⁠02 Price $2 / $12 Length 1.05M
6–25GPT-5.6 SolOpenAIRating 1481 ±701485rank 10–30$4$201.05MOpenAI Released 2026-⁠07 Price $4 / $20 Length 1.05M
6–26Claude Opus 5AnthropicRating 1481 ±7+11491rank 8–22$5$251MAnthropic Released 2026-⁠07 Week change +1 Price $5 / $25 Length 1M
6–26Gemini 3 ProGoogleRating 1481 ±701486rank 10–29——1.05MGoogle Released 2025-⁠11 Price — Length 1.05M
10–31Kimi K3⁠MoonshotRating 1475 ±7−11487rank 8–27$3$151.05MMoonshot Released 2026-⁠07 Week change −1 Price $3 / $15 Length 1.05M
10–30Claude Opus 4.8AnthropicRating 1475 ±601481rank 12–36$5$251MAnthropic Released 2026-⁠05 Price $5 / $25 Length 1M
10–33Gemini 3.6 FlashGoogleRating 1474 ±701483rank 10–35$0.75$3.751.05MGoogle Released 2026-⁠07 Price $0.75 / $3.75 Length 1.05M
10–37Muse Spark 1.3MetaRating 1474 ±9+31496rank 3–15$1.25$4.251.05MMeta Released 2026-⁠09 Week change +3 Price $1.25 / $4.25 Length 1.05M
10–32Gemini 3.5 FlashGoogleRating 1474 ±601477rank 18–41$1.5$91.05MGoogle Released 2026-⁠05 Price $1.5 / $9 Length 1.05M
6–40Claude Sonnet 5.5AnthropicRating 1474 ±13+31478rank 14–44$2$101MAnthropic Released 2026-⁠09 Week change +3 Price $2 / $10 Length 1M
10–33Qwen3.8 MaxAlibabaRating 1474 ±8−21483rank 10–36$2$61MAlibaba Released 2026-⁠08 Week change −2 Price $2 / $6 Length 1M
10–48GPT-6.1 SolOpenAIRating 1469 ±13−91480rank 10–41$2$101.05MOpenAI Released 2026-⁠09 Week change −9 Price $2 / $10 Length 1.05M
11–38GPT-5.5OpenAIRating 1469 ±601482rank 12–36$5$301.05MOpenAI Released 2026-⁠04 Price $5 / $30 Length 1.05M
11–42GLM-5.3⁠Z.aiRating 1468 ±8+21481rank 11–38$1.4$4.41MZ.ai Released 2026-⁠08 Week change +2 Price $1.4 / $4.4 Length 1M
14–39Claude Opus 4.5AnthropicRating 1467 ±501474rank 23–46$5$25200KAnthropic Released 2025-⁠11 Price $5 / $25 Length 200K
11–49GPT-6 AstraOpenAIRating 1465 ±10+11476rank 17–44$10$501.05MOpenAI Released 2026-⁠09 Week change +1 Price $10 / $50 Length 1.05M
14–45GPT-5.4OpenAIRating 1464 ±601475rank 19–42$2.5$151.05MOpenAI Released 2026-⁠03 Price $2.5 / $15 Length 1.05M
14–54Muse SparkMetaRating 1462 ±1101489rank 8–27———Meta Released 2026-⁠04 Price — Length —
15–53Gemini 3 FlashGoogleRating 1461 ±801473rank 23–49——1.05MGoogle Released 2025-⁠12 Price — Length 1.05M
16–54GPT-5.5 InstantOpenAIRating 1460 ±801473rank 20–50———OpenAI Released 2026-⁠05 Price — Length —
19–53GLM-5.2⁠Z.aiRating 1460 ±6+11475rank 19–43$1.4$4.41MZ.ai Released 2026-⁠06 Week change +1 Price $1.4 / $4.4 Length 1M
14–60Muse Spark 1.2MetaRating 1460 ±12−61491rank 6–25$1.25$4.251.05MMeta Released 2026-⁠08 Week change −6 Price $1.25 / $4.25 Length 1.05M
19–54Grok 4.5xAIRating 1458 ±701466rank 33–55$2$6500KxAI Released 2026-⁠07 Price $2 / $6 Length 500K
21–54Muse Spark 1.1MetaRating 1457 ±6−11490rank 8–23$1.25$4.251MMeta Released 2026-⁠07 Week change −1 Price $1.25 / $4.25 Length 1M
11–70Qwen3.7 Max (preview)AlibabaRating 1457 ±2001474rank 15–53————Alibaba Released — Price — Length —
17–62DeepSeek-V4.1 Flash⁠DeepSeekRating 1456 ±1001476rank 17–46——1MDeepSeek Released 2026-⁠09 Price — Length 1M
12–70MiMo-V2.6 Pro⁠XiaomiRating 1456 ±19+11480rank 10–44$0.435$0.871.05MXiaomi Released 2026-⁠09 Week change +1 Price $0.435 / $0.87 Length 1.05M
22–56Claude Sonnet 4.6AnthropicRating 1455 ±601472rank 24–49$3$151MAnthropic Released 2026-⁠02 Price $3 / $15 Length 1M
20–62Qwen3.5 Max (preview)AlibabaRating 1455 ±901465rank 36–60————Alibaba Released — Price — Length —
24–60Claude Sonnet 4.5AnthropicRating 1454 ±501456rank 49–69$3$15200KAnthropic Released 2025-⁠09 Price $3 / $15 Length 200K
19–66DeepSeek-V4 Pro (2026-⁠08)⁠DeepSeekRating 1453 ±1201465rank 30–64$0.66$1.981MDeepSeek Released 2026-⁠08 Price $0.66 / $1.98 Length 1M
How to read this page
  • These ratings exclude the effect of response length and formatting. Differences where the ± ranges overlap are within the margin of error, so about the same.
  • 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