LLM / cheapest LLM API

The cheapest LLM API in 2026

The cheapest LLM API we track is DeepSeek V4 Flash (DeepSeek) at about $17.19/month on our example chatbot workload. Cheapest per class — frontier (flagship-class): Mistral Large 3 ($82); mid: Mistral Small 4 ($29); budget: DeepSeek V4 Flash ($17.19). All 28 models priced by the same engine as the calculator, ranked by cost below.

Prices verified June 2026 · changes logged in the changelog

Every LLM API, ranked by cost

All 28 models we track, cheapest first, on one example workload — chatbot, 100k requests/mo, 2,000 input / 300 output tokens per request, 70% of input cached. Prompt caching is already priced in. The cheapest row is highlighted.

Model$ input /1M$ output /1M$ cached /1M≈ $/mo *
DeepSeek V4 FlashBUDGETDeepSeek $0.14$0.28$0.0028$17.19
Ministral 3 (3B)BUDGETMistral $0.10$0.10$23
Mistral Small 4MIDMistral $0.10$0.30$29
Devstral Small 2BUDGETMistral $0.10$0.30$29
Ministral 3 (8B)BUDGETMistral $0.15$0.15$34.5
Ministral 3 (14B)BUDGETMistral $0.20$0.20$46
GPT-5.4 nanoBUDGETOpenAI $0.20$1.25$0.02$52.3
DeepSeek V4 ProMIDDeepSeek $0.435$0.87$0.003625$52.71
Gemini 3.1 Flash-LiteBUDGETGoogle (Gemini) $0.25$1.50$0.025$63.5
Mistral Large 3FRONTIERMistral $0.50$1.50$0.05$82
CodestralMIDMistral $0.30$0.90$87
Gemini 3 Flash PreviewMIDGoogle (Gemini) $0.50$3$0.05$127
Devstral 2FRONTIERMistral $0.40$2$140
Magistral SmallMIDMistral $0.50$1.50$145
Grok Build 0.1MIDxAI (Grok) $1$2$0.20$148
Grok 4.3FRONTIERxAI (Grok) $1.25$2.50$0.20$178
GPT-5.4 miniMIDOpenAI $0.75$4.50$0.075$190.5
Claude Haiku 4.5BUDGETAnthropic $1$5$0.10$224
Gemini 3.5 FlashMIDGoogle (Gemini) $1.50$9$0.15$381
Gemini 3.1 Pro PreviewFRONTIERGoogle (Gemini) $2$12$0.20$508
Mistral Medium 3.5FRONTIERMistral $1.50$7.50$525
Magistral MediumFRONTIERMistral $2$5$550
GPT-5.4FRONTIEROpenAI $2.50$15$0.25$635
Claude Sonnet 4.6MIDAnthropic $3$15$0.30$672
Claude Opus 4.8FRONTIERAnthropic $5$25$0.50$1,120
GPT-5.5FRONTIEROpenAI $5$30$0.50$1,270
Claude Fable 5FRONTIERAnthropic $10$50$1$2,240
GPT-5.5 ProFRONTIEROpenAI $30$180$11,400

* Same engine as the calculator. Your real number depends on volume, token mix and cache share — tier (frontier / mid / budget) is our editorial class, not a benchmark.

Cheapest for your workload, not ours
Change the volume, token mix and cache share and the calculator re-ranks every model live.
Open calculator
All 28 models → OpenAI pricing → Anthropic pricing → Gemini pricing → DeepSeek pricing → Grok pricing → Mistral pricing → Open calculator →

Frequently asked questions

On our example chatbot workload the cheapest LLM API we track is DeepSeek V4 Flash (DeepSeek) at about $17.19/month. Budget-tier models are almost always the lowest cost, but the cheapest model for you depends on your token mix and how much of the prompt you cache — the calculator re-ranks for your numbers.
Among frontier-class models the lowest cost we track is Mistral Large 3 (Mistral) at about $82/month on the example workload. See the ranked table above for every tier.
Per-token list prices are hard to compare because input, output and cached tokens are billed at different rates. The ≈ $/mo column runs one realistic workload through every model so you can compare them on a single number. Per-1M token rates are in the table too.
Every per-token rate is taken from the provider's official pricing and verified by hand (June 2026). The monthly figure is computed, not quoted. Every change we record lands in the price changelog.