Fieldbook to AIThe plain-English guide to which AI to use

For power users

Advanced table

This is the raw-numbers view for people comfortable with developer terms; for plain-English picks, start from the home page.

Last updated .

Costs here use the prices AI companies charge software developers, so they show relative cost, not your monthly bill. See what ChatGPT, Claude and Gemini charge you.

Click a column heading to sort. 32 models, chosen from 466 listed on OpenRouter.

Glossary. A token is a piece of text, about three-quarters of a word. Input is what you send; output is what the AI writes back; prices are dollars per million tokens. Context is how much the model can read at once. Open weights means the model can be downloaded. The Intelligence Index is from Artificial Analysis as passed along by OpenRouter; Smartness is our 1 to 5 banding of it.
ModelTypeInput $/1M tokensOutput $/1M tokensContext (tokens)Intelligence IndexSmartness (1-5)Released
Claude Opus 5.5
anthropic
Most capable$4.00$20.001M57.652026-09-22
Claude Sonnet 5.5
anthropic
Most capable$2.00$10.001M5652026-09-28
Claude Fable 5.1
anthropic
Most capable$10.00$50.001M53.452026-09-01
GPT-6 Astra
openai
Most capable$10.00*$50.00*1.05M52.752026-09-04
Gemini 4 Argon
google
Most capable$2.00$10.001.05M52.652026-10-13
GPT-6.1 Sol
openai
Most capable$2.00*$10.00*1.05M51.852026-09-29
Muse Spark 1.3
meta
Everyday and cheap$1.25$4.251.05M48.142026-09-02
Grok 4.7
x-ai
Most capable$2.00*$6.00*500K46.442026-09-21
MiMo-V2.6-Pro
xiaomi
Free to downloadopen weights$0.43$0.871.05M46.342026-09-21
GLM 5.3
z-ai
Free to downloadopen weights$0.079$4.801.05M44.842026-08-18
Kimi K3
moonshotai
Free to downloadopen weights$0.30$11.941.05M43.642026-07-16
Claude Haiku 5.5
anthropic
Everyday and cheap$0.10*$0.50*1M43.442026-10-07
Gemini 3.8 Flash
google
Everyday and cheap$0.82$4.131.05M40.942026-09-02
DeepSeek V4.1 Flash
deepseek
Everyday and cheapopen weights$0.15*$0.60*1.05M39.532026-09-10
Mistral Large 4
mistralai
Most capable$0.68$2.091.05M38.432026-10-06
GPT-6 Luna
openai
Everyday and cheap$0.10*$0.50*1.05M38.132026-09-22
MiMo-V2.6-Flash
xiaomi
Everyday and cheapopen weights$0.14$0.281.05M37.932026-09-21
DeepSeek V4 Pro 0813
deepseek
Free to downloadopen weights$0.66*$1.98*1.05M3632026-08-12
Qwen3.8 27B
qwen
Free to downloadopen weights$0.24$2.201M33.732026-08-14
Nemotron 3.5 Super VL
nvidia
Free to downloadopen weights$0.15$0.501.05M26.522026-10-15
Gemma 4 31B
google
Free to downloadopen weights$0.20$0.40262K14.712026-04-02
gpt-oss-120b
openai
Free to downloadopen weights$0.15$0.60131K11.612025-08-05
Sonar Pro
perplexity
Single-purpose$3.00$15.00200K7.612025-03-07
Muse Wonder
meta
Most capable$4.00$20.001.05M––2026-10-12
Command A+
cohere
Most capable$0.30$1.50192K––2026-09-22
Qwen3.8 Flash
qwen
Everyday and cheapopen weights$0.15$0.471M––2026-08-26
Muse Spark Open
meta
Free to downloadopen weights$1.25$4.251.05M––2026-10-15
LFM2.5-24B-A2B
liquid
Free to downloadopen weights$0.050$0.70128K––2026-10-13
Schematron V2 Turbo
inference-net
Single-purposeopen weights$0.030$0.15128K––2026-09-12
Moondream 3.1
moondream
Single-purposeopen weights$0.30$2.5032K––2026-10-15
Nano Banana 2.1
google
Single-purpose$1.50$7.5066K––2026-10-06
Hy-MT2-7B
tencent
Single-purposeopen weights$0.074$0.298K––2026-08-19

Showing the key columns. Open this page on a wider screen to see context size, index score and release date.

Charts

Smarter does not always mean pricier

Among the models rated 5 of 5, a 100-message chat costs from $2.62 to $13.12; for 4 of 5 it runs from $0.13 to $2.54. A higher rating does not always cost more.

Swipe sideways to see the whole chart.

1 of 52 of 53 of 54 of 55 of 5 $0.50$1.00$5.00$10.00 Cost of a 100-message chat (log scale, cheaper to the left) Claude Opus 5.5: ★★★★★ (5 of 5), 100-message chat $5.25Claude Sonnet 5.5, GPT-6.1 Sol, Gemini 4 Argon: ★★★★★ (5 of 5), 100-message chat $2.62Claude Fable 5.1, GPT-6 Astra: ★★★★★ (5 of 5), 100-message chat $13.12Grok 4.7: ★★★★☆ (4 of 5), 100-message chat $2.54Mistral Large 4: ★★★☆☆ (3 of 5), 100-message chat $0.87Claude Haiku 5.5: ★★★★☆ (4 of 5), 100-message chat $0.13GPT-6 Luna: ★★★☆☆ (3 of 5), 100-message chat $0.13Gemini 3.8 Flash: ★★★★☆ (4 of 5), 100-message chat $1.08Muse Spark 1.3: ★★★★☆ (4 of 5), 100-message chat $1.60DeepSeek V4.1 Flash: ★★★☆☆ (3 of 5), 100-message chat $0.19MiMo-V2.6-Flash: ★★★☆☆ (3 of 5), 100-message chat $0.18MiMo-V2.6-Pro: ★★★★☆ (4 of 5), 100-message chat $0.54Kimi K3: ★★★★☆ (4 of 5), 100-message chat $0.61GLM 5.3: ★★★★☆ (4 of 5), 100-message chat $0.19Qwen3.8 27B: ★★★☆☆ (3 of 5), 100-message chat $0.33Nemotron 3.5 Super VL: ★★☆☆☆ (2 of 5), 100-message chat $0.19gpt-oss-120b: ★☆☆☆☆ (1 of 5), 100-message chat $0.19Gemma 4 31B: ★☆☆☆☆ (1 of 5), 100-message chat $0.25DeepSeek V4 Pro 0813: ★★★☆☆ (3 of 5), 100-message chat $0.84Sonar Pro: ★☆☆☆☆ (1 of 5), 100-message chat $3.94 Sonar ProGPT-6 LunaDeepSeek V4.1 FlashClaude Haiku 5.5Gemini 3.8 FlashClaude Sonnet 5.5 · GPT-6.1 Sol · Gemini 4 ArgonClaude Opus 5.5
Show the data as a table
ModelSmartness100-message chat
Claude Sonnet 5.5★★★★★ (5 of 5)$2.62
GPT-6.1 Sol★★★★★ (5 of 5)$2.62
Gemini 4 Argon★★★★★ (5 of 5)$2.62
Claude Opus 5.5★★★★★ (5 of 5)$5.25
Claude Fable 5.1★★★★★ (5 of 5)$13.12
GPT-6 Astra★★★★★ (5 of 5)$13.12
Claude Haiku 5.5★★★★☆ (4 of 5)$0.13
GLM 5.3★★★★☆ (4 of 5)$0.19
MiMo-V2.6-Pro★★★★☆ (4 of 5)$0.54
Kimi K3★★★★☆ (4 of 5)$0.61
Gemini 3.8 Flash★★★★☆ (4 of 5)$1.08
Muse Spark 1.3★★★★☆ (4 of 5)$1.60
Grok 4.7★★★★☆ (4 of 5)$2.54
GPT-6 Luna★★★☆☆ (3 of 5)$0.13
MiMo-V2.6-Flash★★★☆☆ (3 of 5)$0.18
DeepSeek V4.1 Flash★★★☆☆ (3 of 5)$0.19
Qwen3.8 27B★★★☆☆ (3 of 5)$0.33
DeepSeek V4 Pro 0813★★★☆☆ (3 of 5)$0.84
Mistral Large 4★★★☆☆ (3 of 5)$0.87
Nemotron 3.5 Super VL★★☆☆☆ (2 of 5)$0.19
gpt-oss-120b★☆☆☆☆ (1 of 5)$0.19
Gemma 4 31B★☆☆☆☆ (1 of 5)$0.25
Sonar Pro★☆☆☆☆ (1 of 5)$3.94

Filled, larger dots are models we name in our picks. 8 text models with no smartness score are left out. Smartness is our 1 to 5 banding of the Artificial Analysis index (how).

What a 1,000-word essay costs

Across 31 text models, this costs from <1¢ to $0.07.

Bar scale
Show the data as a table
ModelCost of 1,000-word essay
Schematron V2 Turbounder 1 cent
MiMo-V2.6-Flashunder 1 cent
Hy-MT2-7Bunder 1 cent
Gemma 4 31Bunder 1 cent
Claude Haiku 5.5under 1 cent
GPT-6 Lunaunder 1 cent
Qwen3.8 Flashunder 1 cent
Nemotron 3.5 Super VLunder 1 cent
DeepSeek V4.1 Flashunder 1 cent
gpt-oss-120bunder 1 cent
LFM2.5-24B-A2Bunder 1 cent
MiMo-V2.6-Prounder 1 cent
Command A+under 1 cent
DeepSeek V4 Pro 0813under 1 cent
Mistral Large 4under 1 cent
Qwen3.8 27Bunder 1 cent
Moondream 3.1under 1 cent
Gemini 3.8 Flash$0.01 (about 1 cent)
Muse Spark 1.3$0.01 (about 1 cent)
Muse Spark Open$0.01 (about 1 cent)
GLM 5.3$0.01 (about 1 cent)
Grok 4.7$0.01 (about 1 cent)
Claude Sonnet 5.5$0.01 (about 1 cent)
GPT-6.1 Sol$0.01 (about 1 cent)
Gemini 4 Argon$0.01 (about 1 cent)
Kimi K3$0.02 (about 2 cents)
Sonar Pro$0.02 (about 2 cents)
Claude Opus 5.5$0.03 (about 3 cents)
Muse Wonder$0.03 (about 3 cents)
Claude Fable 5.1$0.07 (about 7 cents)
GPT-6 Astra$0.07 (about 7 cents)

Bars use a straight scale: a bar twice as long means twice the cost. All text models we track. Source: OpenRouter list prices and the assumptions on How we work.

What a 100-message chat costs

Across 31 text models, this costs from $0.04 to $13.12.

Bar scale
Show the data as a table
ModelCost of 100 chat messages
Schematron V2 Turbo$0.04 (about 4 cents)
LFM2.5-24B-A2B$0.07 (about 7 cents)
Hy-MT2-7B$0.10 (about 10 cents)
Claude Haiku 5.5$0.13 (about 13 cents)
GPT-6 Luna$0.13 (about 13 cents)
MiMo-V2.6-Flash$0.18 (about 18 cents)
Qwen3.8 Flash$0.19 (about 19 cents)
GLM 5.3$0.19 (about 19 cents)
Nemotron 3.5 Super VL$0.19 (about 19 cents)
DeepSeek V4.1 Flash$0.19 (about 19 cents)
gpt-oss-120b$0.19 (about 19 cents)
Gemma 4 31B$0.25 (about 25 cents)
Qwen3.8 27B$0.33 (about 33 cents)
Command A+$0.39 (about 39 cents)
Moondream 3.1$0.41 (about 41 cents)
MiMo-V2.6-Pro$0.54 (about 54 cents)
Kimi K3$0.61 (about 61 cents)
DeepSeek V4 Pro 0813$0.84 (about 84 cents)
Mistral Large 4$0.87 (about 87 cents)
Gemini 3.8 Flash$1.08
Muse Spark 1.3$1.60
Muse Spark Open$1.60
Grok 4.7$2.54
Claude Sonnet 5.5$2.62
GPT-6.1 Sol$2.62
Gemini 4 Argon$2.62
Sonar Pro$3.94
Claude Opus 5.5$5.25
Muse Wonder$5.25
Claude Fable 5.1$13.12
GPT-6 Astra$13.12

Bars use a straight scale: a bar twice as long means twice the cost. All text models we track. Source: OpenRouter list prices and the assumptions on How we work.

What summarizing a 50-page PDF costs

Across 31 text models, this costs from <1¢ to $0.37.

Bar scale
Show the data as a table
ModelCost of 50-page PDF summary
Schematron V2 Turbounder 1 cent
LFM2.5-24B-A2Bunder 1 cent
Hy-MT2-7Bunder 1 cent
Claude Haiku 5.5under 1 cent
GPT-6 Lunaunder 1 cent
MiMo-V2.6-Flashunder 1 cent
Qwen3.8 Flash$0.01 (about 1 cent)
DeepSeek V4.1 Flash$0.01 (about 1 cent)
Nemotron 3.5 Super VL$0.01 (about 1 cent)
gpt-oss-120b$0.01 (about 1 cent)
GLM 5.3$0.01 (about 1 cent)
Gemma 4 31B$0.01 (about 1 cent)
Qwen3.8 27B$0.01 (about 1 cent)
Command A+$0.01 (about 1 cent)
Moondream 3.1$0.01 (about 1 cent)
MiMo-V2.6-Pro$0.02 (about 2 cents)
Kimi K3$0.02 (about 2 cents)
DeepSeek V4 Pro 0813$0.02 (about 2 cents)
Mistral Large 4$0.02 (about 2 cents)
Gemini 3.8 Flash$0.03 (about 3 cents)
Muse Spark 1.3$0.04 (about 4 cents)
Muse Spark Open$0.04 (about 4 cents)
Grok 4.7$0.07 (about 7 cents)
Claude Sonnet 5.5$0.07 (about 7 cents)
GPT-6.1 Sol$0.07 (about 7 cents)
Gemini 4 Argon$0.07 (about 7 cents)
Sonar Pro$0.11 (about 11 cents)
Claude Opus 5.5$0.15 (about 15 cents)
Muse Wonder$0.15 (about 15 cents)
Claude Fable 5.1$0.37 (about 37 cents)
GPT-6 Astra$0.37 (about 37 cents)

Bars use a straight scale: a bar twice as long means twice the cost. All text models we track. Source: OpenRouter list prices and the assumptions on How we work.

* Price changes with prompt length or day of week; headline price shown. "–" means not reported. Speed is not shown because we do not yet have a verified source.