Token Ledger – 2026-05-15
356 models added, 0 removed, 0 price changes. The largest influx on record reframes the cost landscape. Leading the batch is a 1-trillion-parameter model at sub-dollar rates.
Most cost-impacting addition
inclusionAI: Ring-2.6-1T – $0.075 / 1M input, $0.625 / 1M output, 262k context.
A 1T-parameter dense Mixture-of-Experts model at this price point is unprecedented. For reference, comparable-scale models typically run 5-10× higher. Developers processing high-volume reasoning tasks should test immediately.
Other notable low-cost entries
- IBM: Granite 4.1 8B – $0.05 / 1M input, $0.10 / 1M output, 131k context. Cheapest 8B in the fleet.
- Google: Gemini 3.1 Flash Lite – $0.25 / 1M input, $1.50 / 1M output, 1M context. Largest context-to-cost ratio on a production model.
- Perceptron: Perceptron Mk1 – $0.15 / 1M input, $1.50 / 1M output, 32k context. New entrant at the ultra-budget tier.
- xAI: Grok 4.3 – $1.25 / 1M input, $2.50 / 1M output, 1M context. Lower than Grok 4.2 pricing.
Premium tier
- Anthropic: Claude Opus 4.7 (Fast) – $30 / 1M input, $150 / 1M output, 1M context. Fast variant of Opus.
- OpenAI: GPT Chat Latest – $5 / 1M input, $30 / 1M output, 400k context. New default chat model.
Free models added
Baidu Qianfan CoBuddy, NVIDIA Nemotron 3 Nano Omni, Poolside Laguna XS.2 & M.1, and OpenRouter Owl Alpha are available at zero cost.
All additions bring the platform to 356 total models. No existing model prices changed.
Originally published at The Token Ledger. Subscribe for the daily digest.