COMPUTED FROM THE CATALOG
Cheapest models for chat, by blended price
Text models with at least a 32K context window, ordered by what a typical conversation costs.
How this list is computed
Models that write only text, with a context window of at least 32,000 tokens and a listed price. Sorted by a blended price of three input tokens to one output token, (3 × input + output) ÷ 4 per million tokens, lowest first; ties go to the larger context window, then the name.
321 of the 350 models in the 2026-10-24 snapshot qualify. This is not a quality ranking: SEKAI has no benchmark data.
| # | Model | Provider | Added | Context | Input | Output |
|---|---|---|---|---|---|---|
| 1 | Extract$0.030 blended per million tokens | ScaleDown | 2026-10-13 | 1M | $0.040 | Free |
| 2 | Summarize$0.030 blended per million tokens | ScaleDown | 2026-10-13 | 1M | $0.040 | Free |
| 3 | Ling 3.0 Flash$0.030 blended per million tokens | inclusionAI | 2026-07-23 | 262K | $0.020 | $0.060 |
| 4 | Mistral Nemo$0.030 blended per million tokens | Mistral | 2024-07-19 | 131K | $0.030 | $0.030 |
| 5 | gpt-oss-20b$0.037 blended per million tokens | OpenAI | 2025-08-05 | 131K | $0.020 | $0.090 |
| 6 | Gemma 3 4B$0.042 blended per million tokens | 2025-03-13 | 131K | $0.040 | $0.050 | |
| 7 | Granite 4.0 Micro$0.042 blended per million tokens | IBM | 2025-10-20 | 131K | $0.020 | $0.110 |
| 8 | Nex-N2.5-Mini$0.048 blended per million tokens | Nex AGI | 2026-09-08 | 262K | $0.030 | $0.100 |
| 9 | Qwen3.7 Flash$0.055 blended per million tokens | Qwen | 2026-07-27 | 1M | $0.030 | $0.130 |
| 10 | Llama 3.1 8B Instruct$0.058 blended per million tokens | Meta | 2024-07-23 | 131K | $0.050 | $0.080 |
Questions and answers
Which model comes first in this list?
Extract from ScaleDown: $0.030 blended per million tokens.
How many models meet the criteria?
321 of the 350 models in the 2026-10-24 snapshot.
Which qualifying model has the largest context window?
Grok 4.20 (SpaceXAI), with 2M tokens.
Which qualifying model has the lowest input price?
DeepSeek V4.1 Flash (DeepSeek), at $0.020 per million input tokens.
Answers are computed from the 2026-10-24 catalog snapshot.