Inferse
For vendors
AdvertiseTop spots, featured rows, home sponsor Get listedAdd a product we don't cover yet Claim a listingFree: verify the domain, then correct your page Vendor dashboardNumbers, orders and placements
List your platform
Vendor dashboard

Cheapest Language Models in 2026

Updated

437 on record. Sorted by input price; every figure comes from the OpenRouter model catalogue, with the page it was read from on each entry.

437listed
#ModelContext windowInput / 1MOutput / 1M
201Llama 3.1 70B InstructMetaopen weightstext131K$0.41$0.41
202Mistral Medium 3Mistraltext, image, file131K$0.41$21
203Mistral Medium 3.1Mistraltext, image, file131K$0.41$21
204Qwen3 VL 235B A22B ThinkingQwenopen weightstext, image131K$0.41$41
205UnslopNemo 12BTheDrummeropen weightstext1M$0.41$0.41
206Weaver (alpha)Mancertext8K$0.41$0.751
207GLM 5.2Z.aiopen weightstext1M$0.411$3.991
208ERNIE 4.5 VL 424B A47B Baiduopen weightsimage, text123K$0.421$1.251
209Qwen3.8 27BQwenopen weightstext, image, video1M$0.421$31
210GLM 4.6Z.aiopen weightstext205K$0.431$1.751
211Kimi K2.6MoonshotAIopen weightstext, image262K$0.434$1.828
212MiMo-V2.5-ProXiaomiopen weightstext1.1M$0.435$0.871
213MiMo-V2.6-ProXiaomiopen weightstext, image, video, audio1.1M$0.435$0.871
214GLM 5.3 (batch)Z.aiopen weightstext1M$0.451$21
215Inkling SmallThinking Machinesopen weightstext, image, audio524K$0.451$1.21
216Kimi K2.5MoonshotAIopen weightstext, image262K$0.451$2.251
217Qwen3 235B A22BQwenopen weightstext131K$0.455$1.821
218Claude Haiku 4.5 (batch)Anthropictext, image, file200K$0.51$2.51
219Gemini 3 Flash PreviewGoogletext, image, file, audio, video1M$0.51$31
220GPT-3.5 TurboOpenAItext16K$0.51$1.51
221Mistral Large 3 2512Mistraltext, image, file262K$0.51$1.51
222Nano Banana 2 (Gemini 3.1 Flash Image)Googleimage, text131K$0.51$31
223Nano Banana 2 (Gemini 3.1 Flash Image Preview)Googleimage, text66K$0.51$31
224Nemotron 3 UltraNVIDIAopen weightstext262K$0.51$2.21
225R1 0528DeepSeekopen weightstext164K$0.51$2.151

Context bars share one log scale, 1K to 2M tokens. Prices are what each provider publishes per million tokens; 'varies' is a router that bills the model it picks.

More in Models

All models lists
Longest-context models 466Open-weight models 178Vision language models 295Free models 22Newest models 466Reasoning models 332Models with tool calling 395Image generation models 9