modelgrep

Cheapest Meta Models

Quick answer · Updated June 2026

The cheapest Meta model is Llama 3.1 8B Instruct at $0.020 per million input tokens. Llama 3.2 1B Instruct ($0.027) and Llama 3.2 3B Instruct ($0.051) round out the top three.

$0.020Input /M
11.8Intelligence
145 t/sSpeed
131KContext

AI models ranked by input token price. The most affordable large language model APIs, from budget open-weight models to discounted frontier models.

  1. 1M
    llama-3.1-8b-instruct
    ToolsJSON11.8 intel · 145 t/s · 143ms ttft
    $0.020
    Input /M
  2. 2M
    llama-3.2-1b-instruct
    6.3 intel · 169 t/s · 332ms ttft
    $0.027
    Input /M
  3. 3M
    llama-3.2-3b-instruct
    102 t/s · 223ms ttft · 131K ctx
    $0.051
    Input /M
  4. 4M
    llama-4-scout
    ToolsJSONVision13.5 intel · 130 t/s · 249ms ttft
    $0.100
    Input /M
  5. 5M
    llama-3.3-70b-instruct
    ToolsJSON14.5 intel · 115 t/s · 244ms ttft
    $0.100
    Input /M
  6. 6M
    llama-3-8b-instruct
    6.4 intel · 63 t/s · 660ms ttft
    $0.140
    Input /M
  7. 7M
    llama-4-maverick
    ToolsJSONVision18.4 intel · 72 t/s · 303ms ttft
    $0.150
    Input /M
  8. 8M
    llama-guard-4-12b
    JSONVision18 t/s · 120ms ttft · 164K ctx
    $0.180
    Input /M
  9. 9M
    llama-3.2-11b-vision-instruct
    JSONVision8.7 intel · 35 t/s · 164ms ttft
    $0.345
    Input /M
  10. 10M
    llama-3.1-70b-instruct
    ToolsJSON12.5 intel · 28 t/s · 303ms ttft
    $0.400
    Input /M
  11. 11M
    llama-3-70b-instruct
    JSON8.9 intel · 18 t/s · 1.3s ttft
    $0.510
    Input /M

Frequently asked

What is the cheapest Meta model?

The cheapest Meta model is Llama 3.1 8B Instruct at $0.020 per million input tokens. Llama 3.2 1B Instruct ($0.027) and Llama 3.2 3B Instruct ($0.051) round out the top three.

What's a good alternative to Llama 3.1 8B Instruct?

Llama 3.2 1B Instruct ($0.027) is the closest alternative on this metric, followed by Llama 3.2 3B Instruct ($0.051). See the full ranking above for the tradeoffs.

How many Meta models are there?

modelgrep tracks 13 Meta models with live benchmarks, speed, latency and per-provider pricing, led on intelligence by Llama 4 Maverick. 11 of them qualify for this ranking.

More Meta rankings

All rankings