THE MODELS, SIDE BY SIDE

Which plan you need for each model, what it costs a month and what you get for free, plus how each one scores. Prices come from the labs' own pages: the US price first, the price in Austria under it.

THE SHORT ANSWER

Free
DeepSeek V4.1 Flash in the DeepSeek app · Kimi K3 in the Kimi app
About $20 a month
GPT-6.2 Astra with ChatGPT Plus ($20/month) · Claude Opus 5.5 with Claude Pro ($20/month) · Gemini 4 Neon with Google AI Pro ($19.99/month)
From $100 a month
Claude Fable 5.2 with Claude Max · Gemini 4 Carbon with Google AI Ultra
Which one is best for what? See the task picksAlready paying for one? Check your setup: keep, drop or switch

THE $20 TIER: IN OUR WEEKLY BLIND TEST

The best model each chat app gives you on its standard paid plan, about $20 a month, plus the two strongest free apps. These are the five models in the weekly blind test.

ModelPlan you needPriceFree in the app?Epoch capability scorePeople's score
GPT-6.2 AstraOpenAI · ChatGPTEPOCH: TOP TIERLOWEST PRICE FOR ITS LEVELChatGPT PlusIn Work and Codex, not regular chatChatGPT pricing$20/month€23/month in AustriaNo (Free gets a Luna model)172.8 (169.4–176.9)Too few votes yet
Claude Opus 5.5Anthropic · ClaudeEPOCH #5 OF 258A SIMILAR MODEL IS FREEClaude ProClaude pricing$20/month, or $17/month billed yearlyfrom €15/month (billed yearly) in AustriaNo (Free gets Sonnet)168.7 (165.9–172.0)Too few votes yet
Gemini 4 NeonGoogle · GeminiGoogle AI ProGoogle AI plans$19.99/month€21.99/month in AustriaNo—Too few votes yet
DeepSeek V4.1 FlashDeepSeek · DeepSeek · OPEN WEIGHTSEPOCH #24 OF 258FREE IN ITS APPFree (Expert mode)DeepSeek chat$0Yes159.4 (157.1–161.9)Too few votes yet
Kimi K3Moonshot AI · Kimi · OPEN WEIGHTSEPOCH #26 OF 258FREE IN ITS APPFree (Adagio)Kimi pricing$0Yes, basic access158.6 (156.7–160.8)Too few votes yet

PREMIUM: PRICIER PLANS

Stronger or less limited models that need a more expensive plan. Not in the weekly blind test, which compares what a typical subscriber gets.

ModelPlan you needPriceFree in the app?Epoch capability score
Claude Fable 5.2Anthropic · ClaudeEPOCH: TOP TIER5× THE PRICE OF SIMILAR MODELSClaude MaxClaude pricing$100/monthfrom €90/month in AustriaNo173.4 (170.0–177.2)
Gemini 4 CarbonGoogle · GeminiEPOCH #3 OF 258A SIMILAR MODEL IS FREEGoogle AI UltraGoogle AI plansFrom $99.99/monthfrom €99.99/month (20× tier €219.99) in AustriaNo169.7 (166.6–173.1)
NEW · 60 SECONDS

WHICH AI IS FOR ME?

Four taps. The answer comes from our blind-test votes, or the editor's pick until a task has enough votes.

See which AI wins each task

COST VS CAPABILITY

Higher means more capable on the Epoch Capabilities Index; further right costs more. Each line shows Epoch's 95% range: points whose lines overlap are about equally capable. Updated with every daily forecast.

Showing subscriptions: what you pay per month in each lab's chat app, with every model placed at the cheapest plan that unlocks it. Pricier tiers that only add usage (Kimi Plus, Pro and Max, ChatGPT Pro's higher tiers) aren't separate points.

Subscriptions: monthly price against Epoch score 18 points. Higher is more capable; further right costs more. The same data is in the table below the chart. 150155160165170175180Free$5$10$20$50$100 Subscription price per month (US) Epoch Capabilities Index GPT-6 Luna (ChatGPT): Epoch 156.4 (range 154.6–158.5). free with ChatGPT Free GPT-6.2 Astra (ChatGPT): Epoch 172.8 (range 169.4–176.9). $20/month with ChatGPT Plus GPT-6.1 Sol (ChatGPT): Epoch 168.6 (range 165.7–171.9). $20/month with ChatGPT Plus GPT-6 Astra (ChatGPT): Epoch 169.2 (range 166.4–172.5). $20/month with ChatGPT Plus GPT-6 Sol (ChatGPT): Epoch 163.8 (range 161.4–166.6). $20/month with ChatGPT Plus GPT-5.6 Sol (ChatGPT): Epoch 163.0 (range 160.7–165.6). $20/month with ChatGPT Plus GPT-5.6 Terra (ChatGPT): Epoch 160.8 (range 158.5–163.3). $20/month with ChatGPT Plus GPT-5.6 Luna (ChatGPT): Epoch 156.7 (range 154.9–158.8). $20/month with ChatGPT Plus GPT-6.1 Terra (ChatGPT): Epoch 167.7 (range 164.9–171.0). $100/month with ChatGPT Pro Claude Sonnet 5.5 (Claude): Epoch 166.6 (range 163.9–169.6). free with Claude Free Claude Haiku 5.5 (Claude): Epoch 160.3 (range 158.2–162.7). free with Claude Free Claude Opus 5.5 (Claude): Epoch 168.7 (range 165.9–172.0). $20/month with Claude Pro Claude Fable 5.2 (Claude): Epoch 173.4 (range 170.0–177.2). $100/month with Claude Max Gemini 4 Hydrogen (Gemini): Epoch 166.8 (range 164.0–170.1). $4.99/month with Google AI Plus Gemini 4 Carbon (Gemini): Epoch 169.7 (range 166.6–173.1). $99.99/month with Google AI Ultra DeepSeek V4.1 Flash 1017 (DeepSeek): Epoch 159.4 (range 157.1–161.9). free with DeepSeek Free (Expert mode) Kimi K3 (Kimi): Epoch 158.6 (range 156.7–160.8). free with Kimi Free (Adagio) Kimi K3.1 Pro (Kimi): Epoch 161.2 (range 158.9–163.7). $39/month with Kimi Pro Subscriptions: monthly price against Epoch score 18 points. Higher is more capable; further right costs more. The same data is in the table below the chart. 150155160165170175180Free$5$20$50$100 Subscription price per month (US) Epoch Capabilities Index

Not shown: ChatGPT Business (the lab doesn't say which model version); Gemini Free (the lab doesn't say which model version); GPT-6.2 Sol (in ChatGPT Plus; Epoch hasn't scored it yet); Gemini 4 Neon (in Google AI Pro; Epoch hasn't scored it yet).

Chart data as a table
ModelAppCheapest way to get itEpoch score (range)
Claude Fable 5.2Claude$100/month with Claude Max173.4 (170.0–177.2)
GPT-6.2 AstraChatGPT$20/month with ChatGPT Plus172.8 (169.4–176.9)
Gemini 4 CarbonGemini$99.99/month with Google AI Ultra169.7 (166.6–173.1)
GPT-6 AstraChatGPT$20/month with ChatGPT Plus169.2 (166.4–172.5)
Claude Opus 5.5Claude$20/month with Claude Pro168.7 (165.9–172.0)
GPT-6.1 SolChatGPT$20/month with ChatGPT Plus168.6 (165.7–171.9)
GPT-6.1 TerraChatGPT$100/month with ChatGPT Pro167.7 (164.9–171.0)
Gemini 4 HydrogenGemini$4.99/month with Google AI Plus166.8 (164.0–170.1)
Claude Sonnet 5.5Claudefree with Claude Free166.6 (163.9–169.6)
GPT-6 SolChatGPT$20/month with ChatGPT Plus163.8 (161.4–166.6)
GPT-5.6 SolChatGPT$20/month with ChatGPT Plus163.0 (160.7–165.6)
Kimi K3.1 ProKimi$39/month with Kimi Pro161.2 (158.9–163.7)
GPT-5.6 TerraChatGPT$20/month with ChatGPT Plus160.8 (158.5–163.3)
Claude Haiku 5.5Claudefree with Claude Free160.3 (158.2–162.7)
DeepSeek V4.1 Flash 1017DeepSeekfree with DeepSeek Free (Expert mode)159.4 (157.1–161.9)
Kimi K3Kimifree with Kimi Free (Adagio)158.6 (156.7–160.8)
GPT-5.6 LunaChatGPT$20/month with ChatGPT Plus156.7 (154.9–158.8)
GPT-6 LunaChatGPTfree with ChatGPT Free156.4 (154.6–158.5)

Sources: Epoch AI, Epoch Capabilities Index (CC BY 4.0); plan prices from the labs' pages (see Plans in detail); API prices from each lab's own servers via OpenRouter. Price isn't everything: plans differ in usage limits and features.

THE FRONTIER OVER TIME

Every model the Epoch Capabilities Index scores, placed at its release date. The step line is the best score so far: each step up is a model that took the lead when it came out. Labs leapfrog each other, so a model can lose the top spot within weeks.

The frontier over time, last 12 months 96 models by release date and Epoch score; the step line is the best score so far. The record-setting models are listed in the table below the chart. 135140145150155160165170175180Jan 26MarMayJulSeptNovClaude Opus 4.5 (Anthropic), released Mon 24 Nov 2025: Epoch 151.9 (range 150.3–153.4)DeepSeek-V3.2 (DeepSeek), released Mon 1 Dec 2025: Epoch 146.3 (range 144.7–147.8)GPT-5.2 (OpenAI), released Thu 11 Dec 2025: Epoch 154.5 (range 152.9–156.3)Gemini 3 Flash (Google), released Wed 17 Dec 2025: Epoch 152.3 (range 150.8–154.1)GLM-4.7 (Zhipu (Z.ai)), released Mon 22 Dec 2025: Epoch 143.3 (range 141.4–145.3)Kimi K2.5 (Moonshot AI), released Tue 27 Jan 2026: Epoch 148.8 (range 147.3–150.3)Claude Opus 4.6 (Anthropic), released Thu 5 Feb 2026: Epoch 156.3 (range 154.6–158.3)GLM-5 (Zhipu (Z.ai)), released Wed 11 Feb 2026: Epoch 146.5 (range 144.9–148.1)Qwen3.5 397B-A17B (Alibaba), released Fri 13 Feb 2026: Epoch 148.0 (range 145.9–150.1)Qwen 3.5 Plus (hosted 397B-A17B) (Alibaba), released Mon 16 Feb 2026: Epoch 147.3 (range 145.4–149.2)Claude Sonnet 4.6 (Anthropic), released Tue 17 Feb 2026: Epoch 152.8 (range 151.2–154.5)Grok 4.20 (xAI), released Tue 17 Feb 2026: Epoch 152.3 (range 150.5–154.1)Gemini 3.1 Pro (Google), released Thu 19 Feb 2026: Epoch 156.3 (range 154.5–158.2)Qwen3.5-35B-A3B (Alibaba), released Tue 24 Feb 2026: Epoch 144.1 (range 141.5–146.6)Qwen3.5-9B (Alibaba), released Tue 24 Feb 2026: Epoch 139.5 (range 137.3–141.7)Qwen 3.5 Flash (hosted 35B-A3B) (Alibaba), released Wed 25 Feb 2026: Epoch 143.8 (range 141.9–145.7)Gemini 3.1 Flash-Lite (Google), released Tue 3 Mar 2026: Epoch 145.0 (range 142.8–147.1)GPT-5.4 (OpenAI), released Thu 5 Mar 2026: Epoch 157.9 (range 156.0–160.0)GPT-5.4 Mini (OpenAI), released Tue 17 Mar 2026: Epoch 149.4 (range 147.7–151.1)GPT-5.4 Nano (OpenAI), released Tue 17 Mar 2026: Epoch 146.4 (range 144.7–148.0)Qwen 3.6 Plus (Alibaba), released Tue 31 Mar 2026: Epoch 148.9 (range 146.8–150.9)Gemma 4 31B IT (Google), released Thu 2 Apr 2026: Epoch 144.1 (range 141.7–146.6)Gemma 4 26B A4B (Google), released Thu 2 Apr 2026: Epoch 141.6 (range 139.1–143.9)GLM-5.1 (Zhipu (Z.ai)), released Tue 7 Apr 2026: Epoch 150.7 (range 148.7–152.8)Muse Spark (Meta), released Wed 8 Apr 2026: Epoch 153.2 (range 150.9–155.5)Qwen 3.6 35B-A3B (Alibaba), released Tue 14 Apr 2026: Epoch 144.1 (range 142.2–145.9)Claude Opus 4.7 (Anthropic), released Thu 16 Apr 2026: Epoch 157.5 (range 155.6–159.6)Grok 4.3 Beta (xAI), released Fri 17 Apr 2026: Epoch 149.9 (range 148.0–151.9)Kimi K2.6 (Moonshot AI), released Mon 20 Apr 2026: Epoch 151.6 (range 149.9–153.4)Qwen 3.6 Max (Preview) (Alibaba), released Mon 20 Apr 2026: Epoch 149.9 (range 147.9–151.9)Qwen3.6 27B (Alibaba), released Wed 22 Apr 2026: Epoch 146.7 (range 144.6–148.9)GPT-5.5 (OpenAI), released Thu 23 Apr 2026: Epoch 160.2 (range 158.2–162.4)DeepSeek-V4-Pro (DeepSeek), released Fri 24 Apr 2026: Epoch 150.2 (range 148.4–152.0)DeepSeek-V4-Flash (DeepSeek), released Fri 24 Apr 2026: Epoch 146.9 (range 143.7–150.2)Qwen 3.6 Flash (Alibaba), released Mon 27 Apr 2026: Epoch 143.7 (range 141.8–145.6)Mistral Medium 3.5 (Mistral), released Tue 28 Apr 2026: Epoch 141.6 (range 138.3–144.7)Gemini 3.5 Flash (Google), released Tue 19 May 2026: Epoch 155.3 (range 153.5–157.3)Qwen3.7-Max (Alibaba), released Tue 19 May 2026: Epoch 154.6 (range 152.6–156.8)Claude Opus 4.8 (Anthropic), released Thu 28 May 2026: Epoch 159.9 (range 157.8–162.2)MiniMax-M3 (Other), released Mon 1 Jun 2026: Epoch 147.8 (range 145.7–150.0)Qwen3.7-Plus (Alibaba), released Tue 2 Jun 2026: Epoch 147.9 (range 146.0–150.0)Nemotron 3 Ultra (Other), released Thu 4 Jun 2026: Epoch 147.0 (range 144.9–149.2)Kimi K2.7 Code (Moonshot AI), released Fri 12 Jun 2026: Epoch 151.2 (range 149.2–153.3)GLM-5.2 (Zhipu (Z.ai)), released Tue 16 Jun 2026: Epoch 153.1 (range 151.3–154.9)Claude Sonnet 5 (Anthropic), released Tue 30 Jun 2026: Epoch 156.9 (range 154.9–159.1)Grok 4.5 (xAI), released Wed 8 Jul 2026: Epoch 155.0 (range 153.1–156.9)GPT-5.6 Sol (OpenAI), released Thu 9 Jul 2026: Epoch 163.0 (range 160.7–165.6)GPT-5.6 Terra (OpenAI), released Thu 9 Jul 2026: Epoch 160.8 (range 158.5–163.3)GPT-5.6 Luna (OpenAI), released Thu 9 Jul 2026: Epoch 156.7 (range 154.9–158.8)Muse Spark 1.1 (Meta), released Thu 9 Jul 2026: Epoch 154.9 (range 152.4–157.7)Inkling-Small (Other), released Wed 15 Jul 2026: Epoch 150.8 (range 149.0–152.5)Inkling (Other), released Wed 15 Jul 2026: Epoch 148.4 (range 146.8–150.1)Kimi K3 (Moonshot AI), released Thu 16 Jul 2026: Epoch 158.6 (range 156.7–160.8)Gemini 3.6 Flash (Google), released Tue 21 Jul 2026: Epoch 155.2 (range 153.4–157.3)Gemini 3.5 Flash-Lite (Google), released Tue 21 Jul 2026: Epoch 145.6 (range 144.0–147.2)DeepSeek V4 Flash 0731 (DeepSeek), released Fri 31 Jul 2026: Epoch 154.4 (range 152.6–156.4)Qwen 3.8 Max (Alibaba), released Sun 2 Aug 2026: Epoch 157.6 (range 155.5–159.8)Muse Spark 1.2 (Meta), released Wed 5 Aug 2026: Epoch 155.0 (range 152.7–157.5)Grok 4.6 (xAI), released Wed 12 Aug 2026: Epoch 157.8 (range 155.9–160.0)Gemini 3.7 Flash (Google), released Thu 13 Aug 2026: Epoch 158.4 (range 156.5–160.6)DeepSeek V4 Pro 0813 (DeepSeek), released Thu 13 Aug 2026: Epoch 156.0 (range 154.1–158.1)GLM-5.3 (Zhipu (Z.ai)), released Fri 14 Aug 2026: Epoch 157.1 (range 155.0–159.4)Qwen 3.8 27B (Alibaba), released Fri 14 Aug 2026: Epoch 152.6 (range 150.7–154.7)GLM-5.3-Flash (Zhipu (Z.ai)), released Thu 20 Aug 2026: Epoch 153.8 (range 151.9–155.8)Qwen3.8 Max (0902) (Alibaba), released Tue 1 Sept 2026: Epoch 156.0 (range 153.9–158.3)Gemini 3.8 Flash (Google), released Wed 2 Sept 2026: Epoch 158.8 (range 156.8–161.0)Muse Spark 1.3 (Meta), released Wed 2 Sept 2026: Epoch 158.2 (range 156.2–160.5)DeepSeek V4.1 Flash (DeepSeek), released Wed 9 Sept 2026: Epoch 157.0 (range 155.1–159.2)Grok 4.7 (xAI), released Mon 21 Sept 2026: Epoch 156.4 (range 154.5–158.5)Claude Opus 5.5 (Anthropic), released Tue 22 Sept 2026: Epoch 168.7 (range 165.9–172.0)GPT-6 Sol (OpenAI), released Tue 22 Sept 2026: Epoch 163.8 (range 161.4–166.6)GPT-6 Luna (OpenAI), released Tue 22 Sept 2026: Epoch 156.4 (range 154.6–158.5)Claude Sonnet 5.5 (Anthropic), released Mon 28 Sept 2026: Epoch 166.6 (range 163.9–169.6)GPT-6.1 Sol (OpenAI), released Tue 29 Sept 2026: Epoch 168.6 (range 165.7–171.9)Beam 1.0 (Other), released Mon 5 Oct 2026: Epoch 150.8 (range 148.6–153.1)Mistral Large 4 Preview (Mistral), released Tue 6 Oct 2026: Epoch 152.7 (range 150.8–154.7)Claude Haiku 5.5 (Anthropic), released Wed 7 Oct 2026: Epoch 160.3 (range 158.2–162.7)Muse Wonder (Meta), released Mon 12 Oct 2026: Epoch 161.7 (range 159.4–164.3)Gemini 4 Argon (Google), released Tue 13 Oct 2026: Epoch 166.5 (range 163.9–169.5)Muse Spark Open (Meta), released Thu 15 Oct 2026: Epoch 155.5 (range 153.3–157.8)DeepSeek V4.1 Flash 1017 (DeepSeek), released Sat 17 Oct 2026: Epoch 159.4 (range 157.1–161.9)Kimi K3.1 Pro (Moonshot AI), released Fri 23 Oct 2026: Epoch 161.2 (range 158.9–163.7)GPT-6.1 Terra (OpenAI), released Mon 26 Oct 2026: Epoch 167.7 (range 164.9–171.0)Mistral Large 4 (Mistral), released Thu 29 Oct 2026: Epoch 154.6 (range 151.8–157.4)Gemini 4 Carbon (Google), released Tue 10 Nov 2026: Epoch 169.7 (range 166.6–173.1)Gemini 4 Hydrogen (Google), released Thu 12 Nov 2026: Epoch 166.8 (range 164.0–170.1)GPT-6.2 Astra (OpenAI), released Tue 17 Nov 2026: Epoch 172.8 (range 169.4–176.9)Gemini 3 Pro (Google), released Tue 18 Nov 2025: Epoch 153.0 (range 151.5–154.8). A new top score when released.GPT-5.2 Pro (OpenAI), released Thu 11 Dec 2025: Epoch 157.3 (range 155.1–159.9). A new top score when released.GPT-5.4 Pro (OpenAI), released Thu 5 Mar 2026: Epoch 160.3 (range 157.9–163.0). A new top score when released.GPT-5.5 Pro (OpenAI), released Thu 23 Apr 2026: Epoch 162.8 (range 160.1–165.9). A new top score when released.Claude Fable 5 (Anthropic), released Tue 9 Jun 2026: Epoch 164.2 (range 161.8–166.9). A new top score when released.Claude Opus 5 (Anthropic), released Fri 24 Jul 2026: Epoch 164.7 (range 162.3–167.5). A new top score when released.Claude Fable 5.1 (Anthropic), released Tue 1 Sept 2026: Epoch 167.1 (range 164.4–170.1). A new top score when released.GPT-6 Astra (OpenAI), released Thu 3 Sept 2026: Epoch 169.2 (range 166.4–172.5). A new top score when released.Claude Fable 5.2 (Anthropic), released Mon 9 Nov 2026: Epoch 173.4 (range 170.0–177.2). A new top score when released. Epoch Capabilities Index The frontier over time, last 12 months 96 models by release date and Epoch score; the step line is the best score so far. The record-setting models are listed in the table below the chart. 135140145150155160165170175180Jan 26AprJulOct Epoch Capabilities Index
New top scores in the last 12 months
  • Mon 9 Nov 2026: Claude Fable 5.2 (Anthropic), 173.4
  • Thu 3 Sept 2026: GPT-6 Astra (OpenAI), 169.2
  • Tue 1 Sept 2026: Claude Fable 5.1 (Anthropic), 167.1
  • Fri 24 Jul 2026: Claude Opus 5 (Anthropic), 164.7
  • Tue 9 Jun 2026: Claude Fable 5 (Anthropic), 164.2
  • Thu 23 Apr 2026: GPT-5.5 Pro (OpenAI), 162.8
  • Thu 5 Mar 2026: GPT-5.4 Pro (OpenAI), 160.3
  • Thu 11 Dec 2025: GPT-5.2 Pro (OpenAI), 157.3
  • Tue 18 Nov 2025: Gemini 3 Pro (Google), 153.0

Source: Epoch AI, Epoch Capabilities Index (CC BY 4.0), updated with every daily forecast. Epoch averages public benchmarks; it doesn't measure everything a model is good at.

WHAT THE NUMBERS MEAN

  • Plan you need: the cheapest plan that lets you pick the model in the lab's own chat app.
  • Price: the lab's US price per month, before tax. Under it, the price shown in Austria, including VAT. Prices differ by country and can change without notice, so check the lab's page before you buy.
  • Badges: "Epoch: top tier" means the model is level with Epoch's #1 within Epoch's uncertainty (one's score falls inside the other's range); otherwise "Epoch #" is its position among all 258 models the Epoch Capabilities Index scores. The price badge compares the plan price with every model in a chat plan of similar capability (the same ones as the cost vs capability chart), meaning one model's Epoch score falls inside the other's range: "Lowest price for its level" is within 10% of the cheapest similar model, and "A similar model is free" means one of them costs nothing.
  • Epoch capability score: the Epoch Capabilities Index, which combines many public benchmarks into one scale (higher is better), with its range. Updated Tue 17 Nov. Epoch AI, CC BY 4.0.
  • People's score: our weekly blind test, where people pick between two anonymous answers. 1000 is average; the range shows how sure we are. Preference, not a fact check. How we score

Spotted a mistake? Email corrections@dailyaiweather.com

FOR DEVELOPERS: API PRICES

If you build with these models instead of using the chat apps. Chat-app users don't pay these prices.

ModelAPI price per million tokens, in / outOur test: cost per answer
GPT-6.2 AstraOpenAI$8 / $40—
Claude Opus 5.5Anthropic$4 / $201.8¢ · 295 words on average
Gemini 4 NeonGoogle$3 / $15—
DeepSeek V4.1 FlashDeepSeek$0.15 / $0.60—
Kimi K3Moonshot AI$3 / $152.0¢ · 210 words on average
Claude Fable 5.2Anthropic$8 / $40—
Gemini 4 CarbonGoogle$2 / $10—
  • API price: the lab's own standard price on its servers, via OpenRouter, per million tokens (about 750,000 words) going in and coming out. Updated daily.
  • Cost per answer: what one blind-test answer actually cost us through the API, averaged over every prompt so far, with the average length.