Baseten adds DeepSeek V4.1 Flash 1017 at $0.30/$1.20 per 1M tokens

WHY IT MATTERS

Model-hosting platforms like Baseten compete less on subscription tiers and more on how fast and cheaply they can expose new models. DeepSeek V4.1 Flash 1017 extends a fast pace of catalog additions this week — Nemotron on Oct 16, Reflection Beam and Meta Muse Spark Open on Oct 17 — well ahead of Baseten's baseline cadence of about every 17 days. That speed is itself a form of pricing pressure on rivals racing to list the same models at competitive per-token rates. Watch whether Baseten's near-daily additions settle back into its slower historical cadence, or whether this marks a lasting shift to faster, more frequent catalog updates as new open models ship.

BY THE NUMBERS

• New DeepSeek V4.1 Flash 1017 row: $0.30 input / $0.007 cached / $1.20 output per 1M tokens
• New Model APIs additions this week: Nemotron (Oct 16), Reflection Beam & Meta Muse Spark (Oct 17)
• 6 real pricing changes in the last 90 days, average cadence ~17 days between changes
• Credits/usage billing touched 201 of 1,727 tracked changes (11.6%) across 153 companies in 180 days

THEIR RATIONALE

Baseten isn't framing these moves as pricing changes. Across its 6 real changes in the last 90 days (about one every 17 days), 4 events were model or feature additions, 2 were price cuts and 1 was a feature change, and the most recent moves, adding Nemotron 3.5 Super VL on Oct 16 and Reflection Beam / Meta Muse Spark Open on Oct 17, are additions rather than repricing. Right now the company is prioritizing catalog breadth over touching its existing rates.

THE SIGNAL

Baseten's move sits inside a broader usage-based pricing churn: 201 of the 1,727 pricing changes tracked across the index in the last 180 days (11.6%) touched credits or usage billing, spanning 153 companies. All six of Baseten's real changes since July fall in that same credits and usage bucket. Elsewhere in AI infrastructure, Anthropic gave Max subscribers monthly API credits worth up to $200 on Oct 12, and ElevenLabs pulled its 3x-credits promo for Creator+ after it ended Oct 12. For a model-hosting marketplace like Baseten, the real competitive lever isn't the plan page, it's the per-token rate card, and that card has now changed three days in a row.

IN CONTEXT

Baseten has kept expanding its Model APIs catalog this week, adding Nemotron 3.5 Super VL on Oct 16, then Reflection Beam and Meta Muse Spark Open on Oct 17, and now DeepSeek V4.1 Flash 1017 — well ahead of its usual ~17-day gap between changes. The new row matches the existing DeepSeek V4.1 Flash row's prices, including the $0.007 cached-input rate it reached when Baseten cut that price 77% in September. That mirrors a wider usage-metered pricing shift: Anthropic added monthly API credits for Max subscribers worth up to $200, and ElevenLabs pulled its 3x-credits promo for Creator+, both the same week. For an inference marketplace, a fast-moving catalog is itself a pricing lever — competition is less about the plan page than how quickly new models show up on the rate card.