Skip to content
PricingSaaS
Sign in
Back
Pricing briefing
BBaseten
Baseten
PACKAGING CHANGE

Baseten added Nvidia's Nemotron Super VL to its Model APIs

By , Co-founder, PricingSaaS

Published

Baseten added NVIDIA's Nemotron 3.5 Super VL model to its API pricing list.

See before and after

Baseten’s pricing page · Change detected

What changed

Baseten added NVIDIA's Nemotron 3.5 Super VL model to its Model APIs pricing list, per Baseten's pricing page as of Oct 16, 2026 — the day Pulse detected the change. The new model is priced at $0.20 per 1M input tokens, $0.04 per 1M cached input tokens, and $0.50 per 1M output tokens, according to Baseten's pricing page as of Oct 16, 2026.

Why it matters

For a product leader evaluating model APIs, the cached-input rate matters as much as the headline price: Nemotron 3.5 Super VL's deep discount on cached tokens relative to standard input rewards workloads with repeated context — system prompts, long documents, multi-turn chats — far more than isolated calls.

If you route traffic across multiple model providers, discounts like this are worth modeling into your own cost-per-session math before committing volume to any single vendor. Watch whether Baseten keeps adding vision-capable models at similarly steep cache discounts — that would signal it's competing on throughput economics, not just headline rate, for multimodal workloads.

In context

This is Baseten's fourth real pricing change since Pulse began tracking its page on Jul 20, 2026, following a 77% cut to DeepSeek V4.1 Flash's cached input price, to $0.007 per 1M tokens, on Sep 22, 2026, and earlier additions like Kimi K3 on Jul 28, 2026 and GLM-5.2's cache-price cut on Jul 25, 2026. This pattern suggests Baseten is actively repricing as new models from partners like NVIDIA become available, rather than setting its catalog once. Usage-based and credit pricing moves are common elsewhere in AI tooling right now: ElevenLabs pulled its Eleven v4 3x-credits promo for Creator+, per Pulse's Oct 14, 2026 record, and Mintlify swapped free credits for a 14-day Pro trial with 5,000 credits on Oct 13, 2026, per Pulse. The sharper read for Baseten specifically: cache discounts, not headline input rates, look like its main lever for competing on cost per workload.

Common questions

How should we estimate our own cost with Nemotron 3.5 Super VL?

Multiply Baseten's per-1M-token input, cached-input and output rates by your expected monthly token volume, weighting cached versus fresh input separately since the cache discount is steep — check your own cache-hit ratio before comparing this model against alternatives.

How big are the usage-based perks other AI vendors are offering right now?

Pulse tracked Anthropic giving Max subscribers monthly API credits worth up to $200 on Oct 12, 2026, and Airbyte cutting its free trial from 30 to 14 days while keeping 400 credits the same day — both show vendors leaning on credits and trial length, not just per-token rates, to compete.

Sources