Baseten cuts Reflection Beam rates up to 72%, four days after launch
Baseten cut per-1M-token rates for its Reflection Beam Model API: input from $1.30 to $0.40, cached input from $0.13 to $0.04, and output from $4.30 to $1.20 - up to 72% off the launch rates.
Baseten’s pricing page · Change Detected
WHY IT MATTERS
For buyers evaluating Baseten's Model APIs, the lesson is that launch-day rates aren't necessarily the floor - Reflection Beam's price has already moved since it was added to the catalog. Watch whether this cut holds or Baseten keeps adjusting individual model rates as usage and competitor pricing shift; the company has been changing its pricing roughly every 15 days since Jul 20, 2026, and this catalog looks likely to keep moving.
BY THE NUMBERS
- Reflection Beam launched Oct 17, 2026 at $1.30 input, $4.30 output per 1M tokens
- Input now $0.40, cache input $0.04, output $1.20 per 1M tokens
- 7 pricing changes in the last 90 days; average gap ≈ 15 days
- 241 credits/usage-billing changes across 175 companies in the index in 180 days
THEIR RATIONALE
Baseten doesn't publish a rationale for the cut. It is the company's third price cut since it was first tracked, and it follows the same pattern as DeepSeek V4.1 Flash, whose cached-input price Baseten cut 77% to $0.007 per 1M tokens on Sep 22, 2026. Baseten has changed its pricing roughly every 15 days since Jul 20, 2026, mostly by adding models and adjusting per-model rates.
THE SIGNAL
This sits inside a broader wave: in the last 180 days, 241 pricing changes across 175 companies in the index touched credits and usage billing, 12% of all tracked changes, with companies like Pika, Ideogram, Mintlify, Invideo, Magnific and Lumaai among the most frequent movers on this pattern. For API-based inference pricing specifically, cuts like this reflect how quickly per-token rates get renegotiated once a model ships: Baseten added Reflection Beam on Oct 17, 2026 at $1.30/$0.13/$4.30 per 1M tokens, and within days revised it down to $0.40/$0.04/$1.20 - rapid repricing that is becoming routine as inference providers compete on raw token cost.
IN CONTEXT
Baseten added Reflection Beam to its Model APIs lineup on Oct 17, 2026 at $1.30 input, $0.13 cached input and $4.30 output per 1M tokens, alongside Meta Muse Spark Open. Those launch rates are already gone: this diff cuts them to $0.40, $0.04 and $1.20. Baseten made a similar move on DeepSeek V4.1 Flash, cutting its cached-input price 77% to $0.007 per 1M tokens on Sep 22, 2026, and this cut lands 3 days after Baseten added DeepSeek V4.1 Flash 1017 at $0.30 input, $1.20 output. Usage-based pricing is common across the index: 241 changes across 175 companies touched credits and usage billing in the last 180 days.