Models·Model board

Cheapest Language Models in 2026

This board covers language models listed under “Cheapest language models.” Compare each provider’s published price per million tokens, the exact model name, maker and release date before choosing.

The page includes a signal score out of 10, plus marks for a public API, open-source code and a free tier. Entries include models such as DeepSeek V4 Flash Latest, Granite 4.0 Micro and gpt-oss-20b.

• 437 models
• Checked 16 Oct 2026

1. #ModelInputOutputContext
2. 1
   D

   DeepSeek V4 Flash LatestDeepSeek · reasoning
   $0.009in / 1M $0.81out / 1M 1.05M context
   →
3. 2
   D

   DeepSeek V4 Flash 0731DeepSeek · reasoning
   $0.011in / 1M $1.28out / 1M 1.05M context
   →
4. 3
   D

   DeepSeek Flash LatestDeepSeek · reasoning · vision
   $0.015in / 1M $0.68out / 1M 1.05M context
   →
5. 4
   I

   Granite 4.0 MicroIBM
   $0.017in / 1M $0.11out / 1M 131K context
   →
6. 5
   O

   gpt-oss-20bOpenAI · reasoning
   $0.018in / 1M $0.090out / 1M 131K context
   →
7. 6
   M

   Mistral NemoMistral
   $0.019in / 1M $0.030out / 1M 131K context
   →
8. 7
   I

   Ling 3.0 FlashinclusionAI · reasoning
   $0.021in / 1M $0.063out / 1M 262K context
   →
9. 8
   I

   Ling 3.0 Flash VLinclusionAI · reasoning · vision
   $0.021in / 1M $0.062out / 1M 262K context
   →
10. 9
    O

    gpt-oss-20b (batch)OpenAI · reasoning
    $0.024in / 1M $0.11out / 1M 131K context
    →
11. 10
    O

    GPT-5 Nano (batch)OpenAI · reasoning · vision
    $0.025in / 1M $0.20out / 1M 400K context
    →
12. 11
    NA

    Nex-N2.5-MiniNex AGI · reasoning · vision
    $0.025in / 1M $0.10out / 1M 262K context
    →
13. 12
    ZA

    GLM Flash LatestZ.ai · reasoning · vision
    $0.026in / 1M $0.93out / 1M 1.05M context
    →
14. 13
    M

    Llama 3.2 1B InstructMeta
    $0.027in / 1M $0.20out / 1M 60K context
    →
15. 14
    D

    DeepSeek V4 Flash 0423DeepSeek · reasoning
    $0.028in / 1M $0.056out / 1M 1.05M context
    →
16. 15
    O

    gpt-oss-120b (batch)OpenAI · reasoning
    $0.030in / 1M $0.14out / 1M 131K context
    →
17. 16
    Q

    Qwen3.7 FlashQwen · reasoning · vision
    $0.030in / 1M $0.13out / 1M 1M context
    →
18. 17
    IN

    Schematron V2 TurboInference.net
    $0.030in / 1M $0.15out / 1M 128K context
    →
19. 18
    A

    Nova Micro 1.0Amazon
    $0.035in / 1M $0.14out / 1M 128K context
    →
20. 19
    O

    gpt-oss-120bOpenAI · reasoning
    $0.037in / 1M $0.17out / 1M 131K context
    →
21. 20
    C

    Command R7B (12-2024)Cohere
    $0.037in / 1M $0.15out / 1M 128K context
    →
22. 21
    S

    Llama 3 8B LunarisSao10K
    $0.040in / 1M $0.050out / 1M 8K context
    →
23. 22
    I

    Mercury 2.5Inception · reasoning
    $0.040in / 1M $0.15out / 1M 260K context
    →
24. 23
    I

    Ling 3.0 Flash FininclusionAI · reasoning
    $0.042in / 1M $0.12out / 1M 262K context
    →
25. 24
    T

    Hy-MT2-1.8BTencent
    $0.044in / 1M $0.18out / 1M 8K context
    →
26. 25
    Q

    Qwen3 30B A3B Instruct 2507Qwen
    $0.048in / 1M $0.19out / 1M 262K context
    →

How to choose

Check the exact model name

Several entries use similar names, including DeepSeek V4 Flash Latest, DeepSeek V4 Flash 0731 and DeepSeek Flash Latest. Treat each name as a separate model entry.

OpenAI’s list also separates gpt-oss-20b from gpt-oss-20b (batch), and GPT-5 Nano (batch).

Compare makers and families

The board identifies each maker and family. Makers include DeepSeek, IBM, OpenAI, Mistral, inclusionAI, Nex AGI and Z.ai.

Family labels such as deepseek, ibm-granite and openai help distinguish models that come from the same maker.

Review release dates

Release dates range from Mistral Nemo on 2024-07-19 to several 2026 entries. Checking the date helps you distinguish older versions from newer ones with similar names.

Check access details

Each listed model has an official-site link on OpenRouter. Use the exact link for the model or batch variant you are considering.

Before choosing, compare the published price per million tokens shown for the entry and confirm whether its API, open-source and free-tier marks fit your needs.

Questions

How is this language-model board ordered?

It is sorted by the published price per million tokens, using each provider’s stated figure.

Why are batch versions listed separately?

gpt-oss-20b (batch) and GPT-5 Nano (batch) are separate entries from their non-batch or other model names.

Which makers appear on the list?

The makers shown are DeepSeek, IBM, OpenAI, Mistral, inclusionAI, Nex AGI and Z.ai.

Where can I find each model’s official page?

Every entry includes an OpenRouter URL for that exact model, family label and version.

Why check release dates?

Release dates separate older entries, such as Mistral Nemo from 2024, from models released in 2025 and 2026.

Prices are each provider’s published price per million tokens. How we rank · Disclosure