Skip to content
Search

Research / Compare

Compare AI

Which AI model leads capability, votes, skills, price and value? Open-weight and closed models are ranked together, with every number named and dated.Which AI model is number one? Open and closed models, ranked from named, dated sources.

Data refreshed through 14 Nov 2026 · Sources: Epoch AI, LMArena and OpenRouter

How rankings work

Scores come from named, dated sources. Models without a confident match are left out rather than estimated; API prices are per 1M tokens and do not represent consumer subscriptions. “Downloadable weights” means the model parameters are public; it does not necessarily mean the training data, code or licence are open.

Overall leaderClaude Fable 5.2View ranking →People’s favouriteGemini 4 CarbonView ranking →Lowest API priceGLM-5.3-FlashView ranking →
Show

Downloadable weights: the model parameters are public; licences vary. API / product only: the weights are not public. “Downloadable” does not necessarily mean the training data or code are open.

Overall capability

Smartest models

Epoch Capabilities Index · higher is better · line shows uncertainty

Epoch AI combines many benchmark results into one capability score per model. The dot is the score; the line is Epoch’s 90% confidence interval — where lines overlap, the order between those models is not certain.

  1. Rank 1Claude Fable 5.2API / product only (API access)Anthropic · Nov 2026173.4 (90% confidence interval 169.9 to 177.5)
  2. Rank 2Gemini 4 CarbonAPI / product only (Unreleased)Google DeepMind · Oct 2026169.8 (90% confidence interval 166.5 to 173.5)
  3. Rank 3GPT-6 AstraAPI / product only (API access)OpenAI · Sep 2026169.0 (90% confidence interval 166.1 to 172.3)
  4. Rank 4Claude Opus 5.5API / product only (API access)Anthropic · Sep 2026168.5 (90% confidence interval 165.6 to 171.8)
  5. Rank 5GPT-6.1 SolAPI / product only (API access)OpenAI · Sep 2026168.5 (90% confidence interval 165.5 to 171.8)

Source: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks (opens in a new tab)’ [online resource]. Data as of 14 Nov 2026. Licensed under CC BY 4.0 (opens in a new tab).Source: Epoch AI (opens in a new tab), CC BY 4.0 (opens in a new tab) · 14 Nov 2026

Head-to-head votes

People’s favourite

LMArena text arena score · higher is better · line shows uncertainty

People ask two anonymous models the same question and vote for the better answer; LMArena turns millions of votes into a rating (text arena, overall category, adjusted for answer length and style). Each model appears once, at its highest-rated variant. Open or closed follows the licence LMArena lists.

  1. Rank 1Gemini 4 CarbonAPI / product only (Licence: Proprietary)Google DeepMind · 4,273 votes · variant (High)1,526 (confidence interval 1,517 to 1,536)
  2. Rank 2Gemini 4 ArgonAPI / product only (Licence: Proprietary)Google DeepMind · 20,779 votes · variant (High)1,522 (confidence interval 1,516 to 1,527)
  3. Rank 3Muse WonderAPI / product only (Licence: Proprietary)Meta AI · 12,696 votes · variant (Max)1,511 (confidence interval 1,505 to 1,517)
  4. Rank 4Claude Opus 5.5API / product only (Licence: Proprietary)Anthropic · 15,793 votes · variant (High)1,507 (confidence interval 1,501 to 1,513)
  5. Rank 5Claude Fable 5API / product only (Licence: Proprietary)Anthropic · 45,469 votes · variant (High)1,504 (confidence interval 1,500 to 1,508)

Source: LMArena leaderboard dataset (opens in a new tab), CC BY 4.0 (opens in a new tab). Leaderboard published 11 Nov 2026.Source: LMArena (opens in a new tab), CC BY 4.0 (opens in a new tab) · 11 Nov 2026

Benchmarks run by Epoch AI

Best by skill

Share of problems solved · higher is better

Each model’s best run across the settings Epoch tested (for example, reasoning effort). Runs that fell back to another model are excluded. Only benchmarks Epoch runs itself are used.

SWE-bench Verified (Epoch run) (opens in a new tab) · top 5 of 32 models

  1. Rank 1Claude Opus 4.7API / product only (API access)Anthropic · Apr 202683.5%
  2. Rank 2GPT-5.5API / product only (API access)OpenAI · Apr 202680.6%
  3. Rank 3Gemini 3.5 FlashAPI / product only (API access)Google DeepMind · May 202679.3%
  4. Rank 4Claude Opus 4.6API / product only (API access)Anthropic · Feb 202678.7%
  5. Rank 5GLM-5.2Downloadable weights (Open weights (unrestricted))Z.ai (Zhipu AI) · Jun 202678.7%

Source: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks (opens in a new tab)’ [online resource]. Data as of 14 Nov 2026. Licensed under CC BY 4.0 (opens in a new tab).Source: Epoch AI (opens in a new tab), CC BY 4.0 (opens in a new tab) · 14 Nov 2026

API economics

Cheapest to run

USD per 1M tokens · lower is cheaper

Models in Epoch’s index released in the 12 months to 14 Nov 2026, ranked by blended API price: (3 × input price + 1 × output price) ÷ 4, a typical chat mix of three input tokens per output token. Input and output prices are shown on each row. API prices are not consumer subscription prices.

  1. Rank 1GLM-5.3-FlashDownloadable weights (Open weights (unrestricted))Z.ai (Zhipu AI) · Aug 2026 · in $0.04 · out $0.078$0.05
  2. Rank 2DeepSeek V4.1 FlashDownloadable weights (Open weights (unrestricted))DeepSeek · Sep 2026 · in $0.051 · out $0.20$0.089
  3. Rank 3Gemma 4 26B A4BDownloadable weights (Open weights (unrestricted))Google DeepMind · Apr 2026 · in $0.068 · out $0.23$0.11
  4. Rank 4DeepSeek V4 Flash 0731Downloadable weights (Open weights (unrestricted))DeepSeek · Jul 2026 · in $0.043 · out $0.30$0.11
  5. Rank 5DeepSeek-V4-FlashDownloadable weights (Open weights (unrestricted))DeepSeek · Apr 2026 · in $0.09 · out $0.18$0.11

Prices: OpenRouter models API (opens in a new tab), as of 14 Nov 2026. Models matched to Epoch’s index by name and API id; models without a confident match are left out.Prices: OpenRouter (opens in a new tab) · 14 Nov 2026

Capability per dollar

Best value

Index score against blended price

Up and to the left is better: more capable for less. Same models and price rule as above.

Capabilities Index ↑Blended API price, USD per 1M tokens (log scale) →

Value frontier: nothing else is both cheaper and smarter

  1. GLM-5.3-FlashDownloadable weights$0.05 / 1M · index 153.8
  2. DeepSeek V4.1 FlashDownloadable weights$0.089 / 1M · index 156.7
  3. DeepSeek V4.1 Flash 1017Downloadable weights$0.15 / 1M · index 159.4
  4. Claude Haiku 5.5API / product only$0.20 / 1M · index 160.4
  5. GPT-6.1 TerraAPI / product only$2.00 / 1M · index 167.7
  6. Gemini 4 CarbonAPI / product only$4.00 / 1M · index 169.8
  7. Claude Fable 5.2API / product only$16.00 / 1M · index 173.4
Table: all 86 models in this chart, with every measure
Models with both an Epoch Capabilities Index score and an OpenRouter price
ModelMakerWeightsReleasedIndexArena scoreBlended / 1MInput / 1MOutput / 1M
Claude Fable 5.2AnthropicAPI / product onlyNov 2026173.4No data$16.00$8.00$40.00
Gemini 4 CarbonGoogle DeepMindAPI / product onlyOct 2026169.81,526$4.00$2.00$10.00
GPT-6 AstraOpenAIAPI / product onlySep 2026169.01,474$20.00$10.00$50.00
Claude Opus 5.5AnthropicAPI / product onlySep 2026168.51,507$8.00$4.00$20.00
GPT-6.1 SolOpenAIAPI / product onlySep 2026168.51,481$4.00$2.00$10.00
GPT-6.1 TerraOpenAIAPI / product onlyOct 2026167.7No data$2.00$1.00$5.00
Claude Fable 5.1AnthropicAPI / product onlySep 2026167.01,502$20.00$10.00$50.00
Claude Sonnet 5.5AnthropicAPI / product onlySep 2026166.61,473$4.00$2.00$10.00
Gemini 4 ArgonGoogle DeepMindAccess unclearSep 2026166.51,522$4.00$2.00$10.00
Claude Opus 5AnthropicAPI / product onlyJul 2026164.71,490$10.00$5.00$25.00
Claude Fable 5AnthropicAPI / product onlyJun 2026164.11,504$20.00$10.00$50.00
GPT-6 SolOpenAIAPI / product onlySep 2026163.81,457$4.00$2.00$10.00
GPT-5.6 SolOpenAIAPI / product onlyJul 2026163.01,485$8.00$4.00$20.00
GPT-5.5 ProOpenAIAPI / product onlyApr 2026162.7No data$67.50$30.00$180
Muse WonderMeta AIAPI / product onlyOct 2026161.71,511$8.00$4.00$20.00
Kimi K3.1 ProMoonshotDownloadable weightsOct 2026161.41,485$4.00$2.00$10.00
GPT-5.6 TerraOpenAIAPI / product onlyJul 2026160.71,468$4.50$2.00$12.00
Claude Haiku 5.5AnthropicAPI / product onlyOct 2026160.41,421$0.20$0.10$0.50
GPT-5.4 ProOpenAIAPI / product onlyMar 2026160.2No data$67.50$30.00$180
GPT-5.5OpenAIAPI / product onlyApr 2026160.11,481$11.25$5.00$30.00
Claude Opus 4.8AnthropicAPI / product onlyMay 2026159.91,482$10.00$5.00$25.00
DeepSeek V4.1 Flash 1017DeepSeekDownloadable weightsOct 2026159.4No data$0.15$0.038$0.50
Gemini 3.8 FlashGoogle DeepMindAPI / product onlySep 2026158.81,497$1.65$0.82$4.13
Kimi K3MoonshotDownloadable weightsJul 2026158.61,487$4.14$0.19$16.00
Gemini 3.7 FlashGoogle DeepMindAPI / product onlyAug 2026158.41,489$1.50$0.75$3.75
Muse Spark 1.3Meta AIAPI / product onlySep 2026158.21,497$2.00$1.25$4.25
GPT-5.4OpenAIAPI / product onlyMar 2026157.81,475$5.63$2.50$15.00
Grok 4.6xAIAPI / product onlyAug 2026157.81,457$3.00$2.00$6.00
Claude Opus 4.7AnthropicAPI / product onlyApr 2026157.41,501$10.00$5.00$25.00
GPT-5.2 ProOpenAIAPI / product onlyDec 2025157.3No data$57.75$21.00$168
GLM-5.3Z.ai (Zhipu AI)Downloadable weightsAug 2026157.11,482$0.96$0.084$3.60
Claude Sonnet 5AnthropicAPI / product onlyJun 2026156.91,461$4.00$2.00$10.00
DeepSeek V4.1 FlashDeepSeekDownloadable weightsSep 2026156.71,475$0.089$0.051$0.20
GPT-5.6 LunaOpenAIAPI / product onlyJul 2026156.71,453$0.45$0.20$1.20
GPT-6 LunaOpenAIAPI / product onlySep 2026156.41,441$0.20$0.10$0.50
Claude Opus 4.6AnthropicAPI / product onlyFeb 2026156.31,503$10.00$5.00$25.00
Gemini 3.1 ProGoogle DeepMindAPI / product onlyFeb 2026156.21,487$4.50$2.00$12.00
Grok 4.7xAIAPI / product onlySep 2026156.21,443$3.00$2.00$6.00
Qwen3.8 Max (0902)AlibabaAPI / product onlySep 2026156.0No data$3.00$2.00$6.00
DeepSeek V4 Pro 0813DeepSeekDownloadable weightsAug 2026155.91,464$1.19$0.11$4.44
Muse Spark OpenMeta AIDownloadable weightsOct 2026155.4No data$2.00$1.25$4.25
Gemini 3.5 FlashGoogle DeepMindAPI / product onlyMay 2026155.31,478$3.38$1.50$9.00
Gemini 3.6 FlashGoogle DeepMindAPI / product onlyJul 2026155.21,483$1.50$0.75$3.75
Muse Spark 1.2Meta AIAPI / product onlyAug 2026155.01,493$2.00$1.25$4.25
Grok 4.5xAIAPI / product onlyJul 2026155.01,466$3.00$2.00$6.00
Muse Spark 1.1Meta AIAPI / product onlyJul 2026154.91,491$2.00$1.25$4.25
Qwen3.7-MaxAlibabaAPI / product onlyMay 2026154.5No data$2.21$1.48$4.42
GPT-5.2OpenAIAPI / product onlyDec 2025154.51,437$4.81$1.75$14.00
DeepSeek V4 Flash 0731DeepSeekDownloadable weightsJul 2026154.4No data$0.11$0.043$0.30
GLM-5.3-FlashZ.ai (Zhipu AI)Downloadable weightsAug 2026153.81,475$0.05$0.04$0.078
GLM-5.2Z.ai (Zhipu AI)Downloadable weightsJun 2026153.01,475$1.71$0.007$6.80
Claude Sonnet 4.6AnthropicAPI / product onlyFeb 2026152.91,472$6.00$3.00$15.00
Mistral Large 4 PreviewMistral AIDownloadable weightsOct 2026152.7No data$1.05$0.70$2.10
Qwen 3.8 27BAlibabaDownloadable weightsAug 2026152.61,434$0.58$0.15$1.88
Gemini 3 FlashGoogle DeepMindAPI / product onlyDec 2025152.31,473$1.13$0.50$3.00
Grok 4.20xAIAPI / product onlyFeb 2026152.3No data$1.56$1.25$2.50
Claude Opus 4.5AnthropicAPI / product onlyNov 2025151.91,474$10.00$5.00$25.00
Kimi K2.6MoonshotDownloadable weightsApr 2026151.61,461$1.34$0.65$3.41
Kimi K2.7 CodeMoonshotDownloadable weightsJun 2026151.2No data$1.32$0.60$3.50
Inkling-SmallThinking MachinesDownloadable weightsJul 2026150.81,405$0.64$0.45$1.20
GLM-5.1Z.ai (Zhipu AI)Downloadable weightsApr 2026150.71,465$2.15$1.40$4.40
DeepSeek-V4-ProDeepSeekDownloadable weightsApr 2026150.21,458$1.05$0.90$1.50
Grok 4.3 BetaxAIAPI / product onlyApr 2026149.9No data$1.56$1.25$2.50
GPT-5.4 MiniOpenAIAPI / product onlyMar 2026149.31,447$1.69$0.75$4.50
Qwen 3.6 PlusAlibabaAPI / product onlyMar 2026148.91,443$0.73$0.33$1.95
Kimi K2.5MoonshotDownloadable weightsJan 2026148.81,450$1.00$0.50$2.50
InklingThinking MachinesDownloadable weightsJul 2026148.41,442$1.76$1.00$4.05
Qwen3.5 397B-A17BAlibabaDownloadable weightsFeb 2026147.91,441$1.35$0.60$3.60
Qwen3.7-PlusAlibabaAPI / product onlyJun 2026147.91,455$0.56$0.32$1.28
MiniMax-M3MiniMaxDownloadable weightsJun 2026147.71,436$0.53$0.30$1.20
Nemotron 3 UltraNvidiaDownloadable weightsJun 2026146.9No data$1.05$0.60$2.40
DeepSeek-V4-FlashDeepSeekDownloadable weightsApr 2026146.81,436$0.11$0.09$0.18
Qwen3.6 27BAlibabaDownloadable weightsApr 2026146.7No data$1.04$0.32$3.20
GLM-5Z.ai (Zhipu AI)Downloadable weightsFeb 2026146.51,457$1.55$1.00$3.20
GPT-5.4 NanoOpenAIAPI / product onlyMar 2026146.41,401$0.46$0.20$1.25
Gemini 3.5 Flash-LiteGoogle DeepMindAPI / product onlyJul 2026145.61,455$0.85$0.30$2.50
Gemini 3.1 Flash-LiteGoogleAPI / product onlyMar 2026145.0No data$0.56$0.25$1.50
Gemma 4 31B ITGoogle DeepMindDownloadable weightsApr 2026144.3No data$0.25$0.20$0.40
Qwen 3.6 35B-A3BAlibabaDownloadable weightsApr 2026144.0No data$0.30$0.10$0.90
Qwen3.5-35B-A3BAlibabaDownloadable weightsFeb 2026144.01,394$0.35$0.14$1.00
Qwen 3.5 Flash (hosted 35B-A3B)AlibabaAccess unclearFeb 2026143.81,396$0.11$0.065$0.26
Qwen 3.6 FlashAlibabaAccess unclearApr 2026143.7No data$0.42$0.19$1.13
GLM-4.7Z.ai (Zhipu AI)Downloadable weightsDec 2025143.31,441$1.00$0.60$2.20
Gemma 4 26B A4BGoogle DeepMindDownloadable weightsApr 2026141.61,437$0.11$0.068$0.23
Mistral Medium 3.5Mistral AIDownloadable weightsApr 2026141.61,426$3.00$1.50$7.50
Qwen3.5-9BAlibabaDownloadable weightsFeb 2026139.3No data$0.19$0.17$0.25

Source: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks (opens in a new tab)’ [online resource]. Data as of 14 Nov 2026. Licensed under CC BY 4.0 (opens in a new tab). Prices: OpenRouter models API (opens in a new tab), as of 14 Nov 2026.Source: Epoch AI (opens in a new tab), CC BY 4.0 (opens in a new tab) · 14 Nov 2026. Prices: OpenRouter (opens in a new tab) · 14 Nov 2026

TodayAI editorial

Published comparisons