The cheapest public models right now are extremely low cost on a blended 3:1 input:output basis, with the floor at $0.001 per 1M tokens. After that, the market clusters hard, with several models sitting at $0.006, $0.008, $0.012, and $0.015 per 1M tokens.
For developers, the practical takeaway is simple: cost is no longer the main differentiator at the low end, because multiple text models are effectively in the same budget band.
The absolute cheapest entry in this dataset is fireworks_ai/accounts/fireworks/models/flux-1-dev-controlnet-union at $0.001 per 1M tokens blended, but the next tier jumps to $0.006 per 1M tokens for fireworks-ai-embedding-up-to-150m. That gap is large in percentage terms, which tells you the ultra-low end is not a smooth price curve, it is a set of sharp steps.
Below $0.01 per 1M tokens, there are three models tied at $0.008 per 1M tokens, perplexity/pplx-decider-flash-v1.2, perplexity/pplx-decider-flash-latest, and baseten/decision-models/baseten-joey-2b-latest. That kind of clustering usually means vendors are competing on product shape, latency, or ecosystem rather than trying to undercut each other on raw token cost.
The next band is still cheap, but less extreme, with fireworks-ai-embedding-150m-to-350m at $0.012 per 1M tokens and nscale/Qwen/Qwen2.5-Coder-3B-Instruct plus nscale/Qwen/Qwen2.5-Coder-7B-Instruct both at $0.015 per 1M tokens. For budget planning, that means you can choose between multiple low-cost options without moving your spend much, so model quality and fit should dominate the decision.
- The floor is $0.001 per 1M tokens blended, but it is an outlier.
- Most of the cheapest public models cluster between $0.006 and $0.015 per 1M tokens.
- There are exact price ties at $0.008 and $0.015 per 1M tokens, so cost alone will not separate them.
- At this price level, pick for task fit, not just token cost.
About this data. Figures are from MyTokenTracker live data (model prices, value for money, open data API), free to use and cite under CC BY 4.0. Snapshot: November 2, 2026.