Together AI expands serverless inference catalog with free and premium models
Together AI added three new models to its Serverless Inference Chat pricing table: PrismML Ternary Bonsai 27B at no cost ($0.00 / $0.00 per 1M tokens), Inkling priced at $1.00 input / $4.05 output per 1M tokens, and LFM2.5-8B-A1B priced at $0.03 input / $0.12 output per 1M tokens.
Together’s pricing page · Change Detected
WHY IT MATTERS
As more inference providers stack free, budget, and premium models within the same serverless catalog, buyers gain leverage to shop within a single vendor's offering rather than across vendors, pressuring the middle of the price range. Watch whether Together continues adding free-tier models as a customer-acquisition lever, and whether the $0.03/$0.12 segment becomes the new commodity floor for small open models.
THEIR RATIONALE
By listing PrismML Ternary Bonsai 27B as fully free alongside two paid options spanning roughly two orders of magnitude in price, Together frames its serverless catalog as optimizing for choice across the cost and performance spectrum: free access for experimentation and lightweight use cases, and metered per-token pricing for higher-capability models.
THE SIGNAL
The raw price list understates what is actually happening: a $0.03/$0.12 model sitting next to a fully free model and a $1.00/$4.05 model shows serverless inference marketplaces increasingly using price segmentation, not just tier structure, to steer usage. Free models absorb experimentation traffic while premium per-token pricing captures willingness-to-pay from production workloads, continuing the broader inference-market pattern of competing on a per-model basis rather than a single platform-wide rate card.