Cheapest AI model for reasoning

Reasoning and chain-of-thought models, sorted cheapest-first by blended $/1M tokens.

Token Price Index › Browse › Reasoning

12 models matched for reasoning tasks. Cheapest right now: Thinking Machines: Inkling Small (free) at Free blended $/M. Sorted cheapest-first, updated daily.

Subscribe — $9/mo
ModelContextInput $/MOutput $/MBlended $/M
Thinking Machines: Inkling Small (free)visionfree
thinkingmachines
1,048,576FreeFreeFree
Thinking Machines: Inkling (free)visionfree
thinkingmachines
1,048,576FreeFreeFree
NVIDIA: Nemotron 3 Nano Omni (free)visionfree
nvidia
256,000FreeFreeFree
Arcee AI: Trinity Large Thinkingcache
arcee-ai
262,144$0.250/M$0.800/M$0.388/M
Qwen: Qwen3 Next 80B A3B Thinking
qwen
262,144$0.150/M$1.20/M$0.412/M
Qwen: Qwen3 VL 30B A3B Thinkingvision
qwen
262,144$0.290/M$1.00/M$0.468/M
Thinking Machines: Inkling Smallvisioncache
thinkingmachines
524,288$0.450/M$1.20/M$0.637/M
DeepSeek: R1 0528cache
deepseek
163,840$0.500/M$2.15/M$0.912/M
MoonshotAI: Kimi K2 Thinkingcache
moonshotai
262,144$0.600/M$2.50/M$1.07/M
Qwen: Qwen3 235B A22B Thinking 2507
qwen
128,000$0.450/M$3.50/M$1.21/M
Thinking Machines: Inklingvisioncache
thinkingmachines
524,288$1.00/M$4.05/M$1.76/M
Perplexity: Sonar Reasoning Provision
perplexity
128,000$2.00/M$8.00/M$3.50/M

Frequently asked questions

How many reasoning models are listed?
There are 12 models matching this filter right now, pulled live from the OpenRouter public API and updated daily.
How often does pricing change?
AI model prices change frequently — sometimes weekly. Token Price Index rebuilds this page daily from the OpenRouter API so you always see current prices. Subscribe for price-drop email alerts.
How is the blended cost calculated?
Blended = (3 × input cost + output cost) / 4. This weights input 3:1 over output, reflecting a typical read-heavy workload. Sort by input or output $/M if your workload differs.
How can I get alerted when a reasoning model price drops?
Token Price Index Pro emails you the moment any model you follow gets cheaper — for $9/mo. No more manually checking provider pricing pages.

Get price-drop alerts for models you use

$9/mo · cancel anytime

Pro subscribers get emailed the instant a model they track drops in price, plus a weekly cheapest-per-task digest and full CSV/JSON export.

Subscribe — $9/mo
Data: OpenRouter public model API, normalized to $/1M tokens. Blended = (3×input + output)/4. Generated 19 Oct 2026 22:00 UTC. Prices change often — confirm on the provider before committing spend.