Baseten adds Nemotron 3.5 Super VL at $0.20 per 1M input tokens

WHY IT MATTERS

For buyers choosing inference providers, a rate card that moves roughly every 28 days is worth rechecking before locking in unit economics. Watch whether Nemotron 3.5 Super VL's listed price holds or gets revised the way DeepSeek V4.1 Flash's cached input was cut 77% just 24 days earlier.

BY THE NUMBERS

• 4 pricing-page changes in the last 90 days, averaging one every 28 days
• Previous change: Sep 22, 2026 — DeepSeek V4.1 Flash cached input cut 77% to $0.007 per 1M tokens
• Tracked since Jul 20, 2026, with 4 real pricing changes logged
• Nemotron 3.5 Super VL: $0.20 input, $0.04 cached input, $0.50 output per 1M tokens

THEIR RATIONALE

Baseten has made 4 pricing changes in the last 90 days — all of them catalog or rate-card tweaks rather than plan restructuring — which says it updates its Model APIs menu continuously as new checkpoints like Nemotron 3.5 Super VL become available, instead of batching additions into periodic overhauls.

THE SIGNAL

This is Baseten's first touch of its pricing page since Sep 22, when it cut DeepSeek V4.1 Flash's cached input price 77% to $0.007 per 1M tokens. Together, that shows a provider treating its rate card as a living document: add new models as they ship, reprice existing ones as the underlying compute costs shift, both inside the same 90-day stretch.

IN CONTEXT

Baseten has made 4 pricing changes since Jul 20, 2026, each one touching its Model APIs rate card rather than plan tiers: it cut DeepSeek V4.1 Flash's cached input price 77% to $0.007 per 1M tokens on Sep 22, added Kimi K3 to the catalog on Jul 28, and cut GLM-5.2's cache input price on Jul 25. Nemotron 3.5 Super VL is the latest addition to that same card, arriving 24 days after the DeepSeek cut and continuing a pattern of small, frequent per-model moves rather than one big restructuring.