AI Model Comparison

Compare 41 frontier AI models on price, context and independent benchmarks.

Prices updated . Benchmarks updated .

41 models

View
Show
Show columns
AI model comparison: price per million tokens, context window and independent benchmark scores, updated Nov 16, 2026.
CompareModelYou send / 1MIt replies / 1MContextGood to knowLaunchedECIGPQAARC-AGI-2HLEAIMESimpleQACritPtArenaArena webTokens/sLatencyIt replies / 1MContext
GPT-6 AstraOpenAI$10.00$50.001.05MZDR · Images · Reasoning · Tools2026-09-04169.195.8%95.0%54.8%100.0%75.6%31.7%1,4411,773n/an/a$50.001.05M
Claude Opus 5.5Anthropic$4.00$20.001MZDR · Images · Reasoning · Tools2026-09-22168.490.6%93.3%n/a100.0%72.2%31.7%1,5161,810n/an/a$20.001M
GPT-6.1 TerraOpenAI$1.00$5.001.05MZDR · Images · Reasoning · Tools2026-10-26168.194.3%n/an/a100.0%57.3%28.0%n/a1,673n/an/a$5.001.05M
GPT-6.1 SolOpenAI$2.00$10.001.05MZDR · Images · Reasoning · Tools2026-09-29167.895.4%94.2%n/a100.0%73.9%31.7%1,4461,751n/an/a$10.001.05M
Claude Sonnet 5.5Anthropic$2.00$10.001MZDR · Images · Reasoning · Tools2026-09-28166.795.6%83.9%46.1%100.0%46.5%31.4%1,4731,770n/an/a$10.001M
Claude Fable 5.1Anthropic$10.00$50.001MNo training · Images · Reasoning · Tools2026-09-01166.4n/a90.0%46.5%100.0%70.8%31.1%1,5141,746n/an/a$50.001M
Gemini 4 ArgonGoogle$2.00$10.001.05MZDR · Images · Reasoning · Tools2026-10-13166.196.4%91.3%n/a100.0%76.0%27.1%1,5291,679n/an/a$10.001.05M
Claude Opus 5Anthropic$5.00$25.001MZDR · Images · Reasoning · Tools2026-07-24164.593.9%90.4%n/a98.9%59.9%29.1%1,5081,691n/an/a$25.001M
GPT-6 SolOpenAI$2.00$10.001.05MZDR · Images · Reasoning · Tools2026-09-22163.894.3%89.6%n/a100.0%60.7%30.9%1,3971,678n/an/a$10.001.05M
Claude Fable 5Anthropic$10.00$50.001MNo training · Images · Reasoning · Tools2026-06-09163.085.9%89.2%n/a100.0%70.7%28.6%1,4921,624n/an/a$50.001M
Muse WonderMeta$4.00$20.001.05MZDR · Images · Reasoning · Tools2026-10-12161.7n/an/an/an/an/a28.3%1,4991,652n/an/a$20.001.05M
Claude Haiku 5.5Anthropic$0.100$0.5001MZDR · Images · Reasoning · Tools2026-10-07160.489.7%61.3%30.1%98.9%23.8%22.6%1,4011,589n/an/a$0.5001M
Claude Opus 4.8Anthropic$5.00$25.001MZDR · Images · Reasoning · Tools2026-05-28159.691.0%72.1%n/a98.3%53.0%20.9%1,4611,556n/an/a$25.001M
GPT-5.5OpenAI$5.00$30.001MZDR · Images · Reasoning · Tools2026-04-23159.490.7%85.0%n/a84.4%63.0%27.1%1,472n/an/an/a$30.001M
DeepSeek V4.1 Flash 1017DeepSeek · Open weights$0.300$1.201MZDR · Images · Reasoning · Tools2026-10-17158.893.2%78.1%n/a99.4%42.5%24.0%n/an/an/an/a$1.201M
Kimi K3Moonshot · Open weights$0.192$16.001.05MZDR · Images · Reasoning · Tools2026-07-16158.393.1%60.4%n/a97.2%50.6%23.4%1,4761,656n/an/a$16.001.05M
Gemini 3.7 FlashGoogle$0.750$3.751.05MZDR · Images · Reasoning · Tools2026-08-13158.294.8%84.6%n/a97.2%69.2%14.3%1,4891,590n/an/a$3.751.05M
Gemini 3.8 FlashGoogle$0.825$4.131.05MZDR · Images · Reasoning · Tools2026-09-02157.695.4%89.2%44.5%98.9%69.7%18.3%1,4981,583n/an/a$4.131.05M
Muse Spark 1.3Meta$1.25$4.251.05MZDR · Images · Reasoning · Tools2026-09-02157.4n/an/an/a99.2%n/a26.0%1,4911,656n/an/a$4.251.05M
Claude Sonnet 5Anthropic$2.00$10.001MZDR · Images · Reasoning · Tools2026-06-30157.090.5%n/an/a94.7%33.7%16.9%1,4431,540n/an/a$10.001M
Grok 4.6xAI$2.00$6.00500KZDR · Images · Reasoning · Tools2026-08-12156.994.0%67.1%n/a99.2%49.3%19.7%1,4291,616n/an/a$6.00500K
GPT-6 LunaOpenAI$0.100$0.5001.05MZDR · Images · Reasoning · Tools2026-09-22156.890.5%59.3%n/a98.9%41.4%19.4%1,3891,579n/an/a$0.5001.05M
Qwen3.8 Max (0902)Qwen$2.00$6.00984KNo training · Images · Reasoning · Tools2026-09-02156.292.3%n/an/a100.0%47.3%17.7%n/a1,676n/an/a$6.00984K
GLM 5.3Z.ai · Open weights$0.070$1.401MZDR · Reasoning · Tools2026-08-14156.290.9%n/an/a91.1%41.0%19.1%1,4741,623n/an/a$1.401M
DeepSeek V4.1 FlashDeepSeek · Open weights$0.140$0.4201MZDR · Images · Reasoning · Tools2026-09-10156.189.8%72.9%n/a98.3%42.9%14.3%n/an/an/an/a$0.4201M
DeepSeek V4 Pro 0813DeepSeek · Open weights$0.111$4.441MZDR · Reasoning · Tools2026-08-13155.991.7%61.3%n/a98.6%52.9%18.0%n/an/an/an/a$4.441M
Gemini 3.1 Pro PreviewGoogle$2.00$12.001.05MZDR · Images · Reasoning · Tools2026-02-19155.294.4%77.1%46.4%95.6%73.5%17.7%1,4801,447n/an/a$12.001.05M
Muse Spark OpenMeta · Open weights$1.25$4.251.05MZDR · Images · Reasoning · Tools2026-10-15155.1n/a57.5%n/an/an/a14.9%n/a1,526n/an/a$4.251.05M
Gemini 3.5 FlashGoogle$1.50$9.001.05MZDR · Images · Reasoning · Tools2026-05-19155.092.8%72.1%n/a95.6%66.2%13.1%1,4821,499n/an/a$9.001.05M
Grok 4.7xAI$2.00$6.00500KZDR · Images · Reasoning · Tools2026-09-21154.692.7%61.4%n/a98.1%56.0%18.0%1,3981,637n/an/a$6.00500K
Grok 4.5xAI$2.00$6.00500KZDR · Images · Reasoning · Tools2026-07-08154.493.4%52.6%n/a97.8%48.3%15.4%1,4481,553n/an/a$6.00500K
Qwen3.7 MaxQwen$1.48$4.421MNo training · Reasoning · Tools2026-05-21154.290.9%n/an/a95.6%55.8%13.4%n/an/an/an/a$4.421M
GLM 5.3 FlashZ.ai · Open weights$0.040$0.0781MZDR · Images · Reasoning · Tools2026-08-26153.290.2%65.8%n/a93.9%n/a15.4%1,4731,607n/an/a$0.0781M
Qwen3.8 27BQwen · Open weights$0.150$1.88262KZDR · Images · Reasoning · Tools2026-08-14151.8n/a42.4%n/an/an/a5.4%1,4381,590n/an/a$1.88262K
Mistral Large 4Mistral$0.700$2.101MZDR · Images · Reasoning · Tools2026-10-06150.688.1%n/an/a98.3%20.0%10.6%1,4301,544n/an/a$2.101M
MiniMax M3MiniMax · Open weights$0.300$1.201MZDR · Images · Reasoning · Tools2026-06-01146.990.9%n/an/a71.1%n/a3.7%1,4311,483n/an/a$1.201M
Gemini 4 CarbonPreviewGoogle$2.00$10.001MZDR · Images · Reasoning · Tools2026-10-28n/an/an/an/an/an/a36.0%1,5401,768n/an/a$10.001M
Particle-1Particle$0.060$0.5001.05MFree · ZDR · Images · Reasoning · Tools2026-10-28n/an/an/an/an/an/an/an/an/an/an/a$0.5001.05M
Kimi K3.1 ProMoonshot · Open weights$0.240$8.401.05MZDR · Images · Reasoning · Tools2026-10-23n/an/an/an/an/an/a24.6%n/a1,695n/an/a$8.401.05M
MiniMax M3.1 FlashMiniMax · Open weights$0.150$0.6001MZDR · Images · Reasoning · Tools2026-10-19n/an/an/an/an/an/an/an/a1,582n/an/a$0.6001M
Qwen3.8 FlashQwen$0.150$0.4701MNo training · Images · Reasoning · Tools2026-08-26n/an/an/an/an/an/an/an/an/an/an/a$0.4701M

Independent scores. ECI is the Epoch Capabilities Index. GPQA is GPQA Diamond. AIME is OTIS Mock AIME. Arena is LMArena Elo. Empty cells mean the current open dataset has no score for that model. Column names link to the benchmark guide.

Tokens per second and latency are via OpenRouter, last 30 minutes, and only when OpenRouter publishes a reading. Many models have none.

What is the best model right now?

GPT-6 Astra leads the frontier models on this page with an Epoch Capabilities Index of 169.1. Claude Opus 5.5 is next at 168.4. GPT-6 Astra costs $50.00 per million output tokens and reads 1.05M of context. Claude Opus 5.5 costs $20.00 per million output tokens and reads 1M of context.

What is the cheapest frontier model?

GLM 5.3 Flash has the lowest published output price among frontier models, at $0.078 per million tokens. Input tokens cost $0.040 per million and the context window is 1M. DeepSeek V4.1 Flash is the next lowest at $0.420 per million output tokens. A lower price is not a quality ranking. Compare the Epoch Capabilities Index before you choose.

Which open-weight models lead?

DeepSeek V4.1 Flash 1017 leads open-weight models that have a published Epoch Capabilities Index, at 158.8. It is from DeepSeek, costs $1.20 per million output tokens, and reads 1M. Kimi K3 from Moonshot follows at 158.3. Open weights means the lab published the weights. The license on the model page says what you can do with them.

FAQ

What is the best model right now?

GPT-6 Astra leads the frontier models on this page with an Epoch Capabilities Index of 169.1. Claude Opus 5.5 is next at 168.4. GPT-6 Astra costs $50.00 per million output tokens and reads 1.05M of context. Claude Opus 5.5 costs $20.00 per million output tokens and reads 1M of context.

What is the cheapest frontier model?

GLM 5.3 Flash has the lowest published output price among frontier models, at $0.078 per million tokens. Input tokens cost $0.040 per million and the context window is 1M. DeepSeek V4.1 Flash is the next lowest at $0.420 per million output tokens. A lower price is not a quality ranking. Compare the Epoch Capabilities Index before you choose.

Which open-weight models lead?

DeepSeek V4.1 Flash 1017 leads open-weight models that have a published Epoch Capabilities Index, at 158.8. It is from DeepSeek, costs $1.20 per million output tokens, and reads 1M. Kimi K3 from Moonshot follows at 158.3. Open weights means the lab published the weights. The license on the model page says what you can do with them.

Where do the benchmark scores come from?

Independent scores, including the Epoch Capabilities Index, come from Epoch AI. Arena Elo comes from the LMArena leaderboard dataset. Both are used under CC BY 4.0. The benchmark file date is 2026-10-31.

How fresh are the prices?

Token prices were fetched Nov 16, 2026, 2:31 AM ET. They are US dollars per million tokens at the current public rate and refresh about hourly. This page does not chart older prices.

How to read this table

Input price is what you pay for the tokens you send. Output price is what you pay for tokens the model writes, and it is usually several times higher. Context window is the most text a model can read in one request. The launch date is the lab's date when we have one, not the day a reseller listed the model.

The default Frontier view is a curated set. All adds the rest of the public catalogue from major labs. Open-weight and Under $1 narrow it further. Tick two to four rows to compare them side by side.

Want a monthly estimate? Use the AI API cost calculator.

Sources and licenses

  • OpenRouter: Prices, context windows, modalities and the model list come from the OpenRouter public API. Data: OpenRouter.
  • Vercel AI Gateway: The no-training flag comes from the Vercel AI Gateway model list.
  • models.dev: Lab launch dates, open-weights flags and knowledge cutoffs come from models.dev. Copyright (c) 2025 models.dev, MIT License.
  • Epoch AI: Independent benchmark scores and the Epoch Capabilities Index (ECI) come from Epoch AI's Benchmarking Hub, used under CC BY 4.0. Epoch AI, Capabilities & benchmarking, epoch.ai/benchmarks. Scores Epoch marks as external keep their original license, and the source is named on each score.
  • LMArena: Arena Elo (text, webdev and vision) comes from the LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), CC BY 4.0, leaderboard date 2026-10-28.
  • Lab-reported scores, when an editor has added them, are the lab's own numbers. Each one links to the source and is labeled lab-reported. They are not used to rank the lists.
  • What a score measures: the benchmark guide.