Cheapest AI models for summarising long documents
What 1,000 summaries cost on each model with a verified API price, as of November 2026. Each summary is 60,000 tokens in and 1,500 out.
| # | Model | Per 1,000 summaries | Price per 1M (in / out) | Source |
|---|---|---|---|---|
| 1 | Ling 3.0 TinyinclusionAI | $0.97 | $0.015 / $0.045 | OpenRouter · Nov 15, 2026 |
| 2 | Llama 3.1 8B InstructMeta | $1.26 | $0.02 / $0.04 | OpenRouter · Nov 15, 2026 |
| 3 | Ling 3.0 FlashinclusionAI | $1.35 | $0.021 / $0.063 | OpenRouter · Nov 15, 2026 |
| 4 | Qwen3.5-9BQwen | $1.37 | $0.022 / $0.032 | OpenRouter · Nov 15, 2026 |
| 5 | Nex-N2.5-MiniNex AGI | $1.65 | $0.025 / $0.10 | OpenRouter · Nov 15, 2026 |
| 6 | Mistral NemoMistral | $1.79 | $0.029 / $0.03 | OpenRouter · Nov 15, 2026 |
| 7 | Llama 3.2 1B InstructMeta | $1.92 | $0.027 / $0.201 | OpenRouter · Nov 15, 2026 |
| 8 | gpt-oss-20bOpenAI | $1.99 | $0.03 / $0.13 | OpenRouter · Nov 15, 2026 |
| 9 | Schematron V2 TurboInference.net | $2.03 | $0.03 / $0.15 | OpenRouter · Nov 15, 2026 |
| 10 | Nova Micro 1.0Amazon | $2.31 | $0.035 / $0.14 | OpenRouter · Nov 15, 2026 |
| 11 | Command R7B (12-2024)Cohere | $2.47 | $0.0375 / $0.15 | OpenRouter · Nov 15, 2026 |
| 12 | Ling 3.1 FlashinclusionAI | $2.90 | $0.045 / $0.135 | OpenRouter · Nov 15, 2026 |
| 13 | Gemma 3 4BGoogle | $3.15 | $0.05 / $0.10 | OpenRouter · Nov 15, 2026 |
| 14 | Gemma 3 12BGoogle | $3.23 | $0.05 / $0.15 | OpenRouter · Nov 15, 2026 |
| 15 | Solar Mini 4Upstage | $3.30 | $0.05 / $0.20 | OpenRouter · Nov 15, 2026 |
| 16 | Schematron V2 SmallInference.net | $3.34 | $0.05 / $0.23 | OpenRouter · Nov 15, 2026 |
| 17 | DeepSeek V4.1 FlashDeepSeek | $3.35 | $0.0507 / $0.203 | OpenRouter · Nov 15, 2026 |
| 18 | Llama 3.2 3B InstructMeta | $3.56 | $0.0509 / $0.335 | OpenRouter · Nov 15, 2026 |
| 19 | GPT-5 NanoOpenAI | $3.60 | $0.05 / $0.40 | OpenRouter · Nov 15, 2026 |
| 20 | Ling 3.0 Flash VLinclusionAI | $3.87 | $0.06 / $0.18 | OpenRouter · Nov 15, 2026 |
Cheapest from each major provider
| # | Model | Per 1,000 summaries | Price per 1M (in / out) | Source |
|---|---|---|---|---|
| 1 | Llama 3.1 8B InstructMeta | $1.26 | $0.02 / $0.04 | OpenRouter · Nov 15, 2026 |
| 2 | Qwen3.5-9BQwen | $1.37 | $0.022 / $0.032 | OpenRouter · Nov 15, 2026 |
| 3 | Mistral NemoMistral | $1.79 | $0.029 / $0.03 | OpenRouter · Nov 15, 2026 |
| 4 | gpt-oss-20bOpenAI | $1.99 | $0.03 / $0.13 | OpenRouter · Nov 15, 2026 |
| 5 | Gemma 3 4BGoogle | $3.15 | $0.05 / $0.10 | OpenRouter · Nov 15, 2026 |
| 6 | DeepSeek V4.1 FlashDeepSeek | $3.35 | $0.0507 / $0.203 | OpenRouter · Nov 15, 2026 |
| 7 | Claude Haiku 5.5Anthropic | $6.75 | $0.10 / $0.50 | Vendor page · Nov 14, 2026 |
| 8 | GLM 5.3 FlashZ.ai | $9.75 | $0.075 / $0.25 | Vendor page · Nov 14, 2026 |
| 9 | MiniMax M2.7MiniMax | $13.86 | $0.21 / $0.84 | OpenRouter · Nov 15, 2026 |
| 10 | Kimi K2 0905Moonshot AI | $39.75 | $0.60 / $2.50 | OpenRouter · Nov 15, 2026 |
| 11 | Grok Build 0.1xAI | $63.00 | $1.00 / $2.00 | Vendor page · Nov 14, 2026 |
How this is worked out
Every model with a price we could verify (on the vendor’s own pricing page, or on OpenRouter for routed models) is priced at the same task size. A model is left out if the prompt does not fit its context window, if its price is not verified, or if it is a free or preview route, which usually has tight rate limits. Cheapest is not best: check task fit on the API prices page before you choose.
Prices change: see recent price changes and announced retirements.