LLM Price Index
As of November 20, 2026
What AI models actually cost today, across every provider you can reach through one MartLLM key. Every figure is read live from the public catalog with no MartLLM fee added — nothing is estimated, and anything that can’t be derived from the live catalog is left out.
As of November 20, 2026, the cheapest LLM API on MartLLM is GPT-OSS 20B at $0.000071 per 1M tokens blended ($0.000018 in / $0.000089 out). The cheapest image API is Lustify v8 (Venice) at $0.0025 per image. The cheapest video API is LongCat Distilled Text-to-Video at $0.0126 / sec. The cheapest music API is MMAudio V2 Text-to-Audio (Venice) at from $0.00023 per second.
Cheapest LLM APIs
Ranked by blended cost per 1M tokens (input weighted 1, output 3), cheapest first.
| # | Model | Input / 1M | Output / 1M | Blended / 1M |
|---|---|---|---|---|
| 1 | GPT-OSS 20B openai/gpt-oss-20b | $0.000018 | $0.000089 | $0.000071 |
| 2 | GPT-OSS 120B openai/gpt-oss-120b | $0.000069 | $0.00030 | $0.00024 |
| 3 | GPT OSS Safeguard 20B openai/gpt-oss-safeguard-20b | $0.000075 | $0.00030 | $0.00024 |
| 4 | Qwen: Qwen3.8 Omni Flash qwen/qwen3.8-omni-flash | $0.00015 | $0.00046 | $0.00038 |
| 5 | Voxtral Mini 3B 2507 mistralai/voxtral-mini-3b-2507 | $0.00040 | $0.00040 | $0.00040 |
| 6 | Gemma 3 4B IT google/gemma-3-4b-it | $0.00046 | $0.00058 | $0.00055 |
| 7 | GLM 4.5 Air z-ai/glm-4.5-air | $0.00013 | $0.00084 | $0.00066 |
| 8 | Gemma 4 E2B google/gemma-4-e2b | $0.00040 | $0.00080 | $0.00070 |
| 9 | Gemma 4 26B A4B IT google/gemma-4-26b-a4b-it | $0.00080 | $0.0010 | $0.00095 |
| 10 | Ministral 3B mistralai/ministral-3-3b-instruct | $0.0010 | $0.0010 | $0.0010 |
Cheapest image, video & music APIs
Cheapest image API $0.0025 per image
Priced per image (or per megapixel).
Lustify v8 (Venice)$0.0025 per image
Z-Image Turbo (Venice)$0.004 per image
Lustify v7 (Venice)$0.0045 per image
Anime WAI (Venice)$0.0045 per image
Venice SD35$0.0045 per image
Cheapest video API $0.0126 / sec
Priced per video, or per second for resolution-based models.
LongCat Distilled Text-to-Video$0.0126 / sec
Grok Imagine Text-to-Video$0.0126 / sec
Longcat Distilled (image-to-video)$0.0126 / sec
Grok Imagine$0.0126 / sec
Grok Imagine Video 1.5$0.0126 / sec
Cheapest music API from $0.00023 per second
Priced per track or job.
MMAudio V2 Text-to-Audio (Venice)from $0.00023 per second
ElevenLabs Sound Effects v2 (Venice)from $0.000575 per second
- Sonilo V1.1 Sound Effectsfrom $0.0014 per second
- Sonilo V1.1 Musicfrom $0.002 per second
Seed Audio 1.0from $0.002 per second
Biggest discounts
How far the best price on the market right now sits below each model’s list price, deepest first.
| Model | % off list | Best price |
|---|---|---|
GPT-OSS 20B openai/gpt-oss-20b | 99.9% | $0.000071 / 1M blended |
Qwen: Qwen3.8 Omni Flash qwen/qwen3.8-omni-flash | 99.9% | $0.00038 / 1M blended |
GLM 4.5 Air z-ai/glm-4.5-air | 99.9% | $0.00066 / 1M blended |
GPT-OSS 120B openai/gpt-oss-120b | 99.9% | $0.00024 / 1M blended |
GPT OSS Safeguard 20B openai/gpt-oss-safeguard-20b | 99.9% | $0.00024 / 1M blended |
GLM 5.3 z-ai/glm-5.3 | 99.8% | $0.0082 / 1M blended |
GLM-5 z-ai/glm-5 | 99.6% | $0.0103 / 1M blended |
Qwen 3.8 27B qwen/qwen-3-8-27b | 99.6% | $0.0071 / 1M blended |
GLM-5.2 z-ai/glm-5.2 | 99.5% | $0.0166 / 1M blended |
Qwen 3 235B A22B Instruct 2507 qwen/qwen3-235b-a22b-2507 | 99.5% | $0.0027 / 1M blended |
Cheapest model in every family
The lowest blended $/1M-token model in each family, cheapest family first — 33 families.
Source: MartLLM public models catalog, read live at https://api.martllm.com/api/v1/public/models. Blended price weights input and output 1:3, with no MartLLM fee added. Media prices are per unit in each model’s own unit. No historical trend is shown because no price history series is published.
LLM Price Index — FAQ
What is the cheapest LLM API right now?
This report ranks every model you can call through MartLLM by blended cost per million tokens (input weighted 1, output 3) and lists the cheapest first, as of November 20, 2026. Prices are read live from the cross-provider catalog with no MartLLM fee added, so the price shown is the price paid.
Where does this pricing data come from?
Every number is derived from MartLLM’s public models catalog (https://api.martllm.com/api/v1/public/models), the same live feed that powers the marketplace. There is no estimation and no historical back-fill: the page states what models cost on the day it is read, and omits any figure it cannot derive from the live catalog.
How often is the LLM Price Index updated?
Continuously. The page is generated per request from the live catalog and dated to the day you read it, so a citation of "the cheapest LLM API as of <date>" reflects the market on that date.