Advertise
Cheapest · 2026-11-19

Cheapest AI model for reasoning

The lowest-priced model we track for reasoning is DeepSeek V4 Pro at $1.32 / $3.96 per 1M tokens. The cheapest Frontier-tier option is Gemini 4 Argon. All 19 models below are sorted by blended list price, cheapest first.

#ModelRegionTierPrice in / outBlended / 1M
1DeepSeek V4 ProDeepSeek · open🇨🇳 CNPro$1.32 / $3.96$1.980src
2Muse Spark OpenMeta · open🇺🇸 USPro$1.25 / $4.25$2.000src
3GPT-6.1 TerraOpenAI🇺🇸 USStrong$1 / $5$2.000src
4Qwen4-MaxAlibaba (Qwen)🇨🇳 CNPro$1.5 / $4.5$2.250src
5Gemini 4 HydrogenGoogle🇺🇸 USPro$1.5 / $7.5$3.000src
6Grok 4.7xAI🇺🇸 USPro$2 / $6$3.000src
7Qwen3.8-MaxAlibaba (Qwen)🇨🇳 CNPro$2 / $6$3.000src
8Gemini 4 ArgonGoogle🇺🇸 USFrontier$2 / $10$4.000src
9Gemini 4 CarbonGoogle🇺🇸 USFrontier$2 / $10$4.000src
10GPT-6.2 SolOpenAI🇺🇸 USPro$2 / $10$4.000src
11GPT-6.1 SolOpenAI🇺🇸 USPro$2 / $10$4.000src
12Kimi K3.1 ProMoonshot AI🇨🇳 CNFrontier$2 / $10$4.000src
13Gemini 4 NeonGoogle🇺🇸 USFrontier$3 / $15$6.000src
14Kimi K3Moonshot AI · open🇨🇳 CNFrontier$3 / $15$6.000src
15Claude Opus 5.5Anthropic🇺🇸 USFrontier$4 / $20$8.000src
16Claude Fable 5.2Anthropic🇺🇸 USFrontier$8 / $40$16.000src
17GPT-6.2 AstraOpenAI🇺🇸 USFrontier$8 / $40$16.000src
18GPT-6 AstraOpenAI🇺🇸 USFrontier$10 / $50$20.000src
19Claude Fable 5.1Anthropic🇺🇸 USFrontier$10 / $50$20.000src

What is the cheapest AI model for reasoning?

Of the 19 models we tag for reasoning, DeepSeek V4 Pro by DeepSeek has the lowest blended list price: $1.32 input / $3.96 output per 1M tokens ($1.980 blended), as of 2026-11-19.

What is the cheapest top-tier model for reasoning?

Among the Frontier-tier models we tag for reasoning, Gemini 4 Argon has the lowest blended list price ($4.00 per 1M tokens).

Is there a cheap open-weight model for reasoning?

Yes. The lowest-priced open-weight model we tag for reasoning is DeepSeek V4 Pro ($1.32 / $3.96 per 1M tokens via its listed API).

How is the blended price calculated?

Blended price = (3 x input price + output price) / 4, per 1M tokens, using list prices from the linked sources. Tiers are editorial, not benchmark scores.

See all 36 models on the leaderboard → Estimate your bill Advertise here