Crusoe: GPT-OSS 20B fine-tuning price $0.40->$2.50 + 2 more changes
What happened
Between the Sep 21, 2026 and Sep 28, 2026 captures of Crusoe's pricing page, Pulse recorded 3 changes: GPT-OSS 20B fine-tuning $0.40->$2.50/1M tok; Nemotron 3.5 Lightning $0.80->$0.40; Qwen3.8 27B $4.00->$2.50; added NVIDIA B200 ($9.65/hr) & AMD MI355x ($7.40/hr) self-serve GPUs. 4 new fine-tuning models added (DeepSeek V4 Flash 0731 QAT, Llama 3.3 70B Instruct, GLM 5.2 FP8 QAT, GLM 5.2 FP8). fine-tuning context windows reduced for DeepSeek V4 Flash 0731 (1M->320K), Gemma 4 31B it (262K->32.7K), GPT-OSS 120B (131K->32.7K), Qwen3.6 35B A3B (256K->131K); prices unchanged. In detail: Fine-tuning model context windows shrunk: DeepSeek V4 Flash 0731 1M->320K, Gemma 4 31B it 262K->32.7K, GPT-OSS 120B 131K->32.7K, Qwen3.6 35B A3B 256K->131K tokens (prices unchanged); Added 4 new Serverless Fine-Tuning models: DeepSeek V4 Flash 0731 QAT ($5.00/1M tok), Llama 3.3 70B Instruct ($6.00/1M tok), GLM 5.2 FP8 QAT ($10.00/1M tok), GLM 5.2 FP8 ($10.00/1M tok); GPT-OSS 20B fine-tuning price $0.40->$2.50/1M tok; Nemotron 3.5 Lightning $0.80->$0.40; Qwen3.8 27B $4.00->$2.50. Added NVIDIA B200 ($9.65/hr) & AMD MI355x ($7.40/hr) self-serve GPUs. Pulse files it as a pricing change (limit decrease, feature added, price increase) with an upward direction. Pulse's latest capture (Nov 2, 2026) lists Managed Inference - DeepSeek V4.1 Flash at $0.01 per 1M cached tokens + $0.20 per 1M output tokens, NVIDIA H200 (141GB) On-Demand at $4.29 per GPU-hr, NVIDIA L40S (48GB) On-Demand at $1.50 per GPU-hr, NVIDIA H100 (80GB) On-Demand at $3.90 per GPU-hr and NVIDIA A100 PCIe (80GB) On-Demand at $2 per GPU-hr.
It lands 63 days after Crusoe's previous change, "Crusoe adds Nemotron 3 Ultra 550B to Serverless Inference at $1/$3.20 per…" (Jul 27, 2026). Before that Pulse recorded "Crusoe: Launched Serverless Fine-Tuning + Self-Serve Deployments" (Jul 20, 2026). Crusoe has been tracked since 2026-06-18.
In the same window Spacelift, Tray.ai, Salad, Scalr and RunPod also changed their Cloud & DevOps pricing pages. Those moves are listed further down, and the Pulse Agent can compare them side by side.
- — · Limit−68%1M320K
- — · FeatureAdded
BeforeNot listed
After4 new Serverless Fine-Tuning models
- — · Price+525%$0.40$2.50
| Plan | Field | Before | After | Change |
|---|---|---|---|---|
| — | Limit | 1M | 320K | −68% |
| — | Feature | Not listed | 4 new Serverless Fine-Tuning models | Added |
| — | Price | $0.40 | $2.50 | +525% |
Values as written on Crusoe's public pricing page in the two captures Pulse compared; where the page states no number, the cells give the wording before and after.
Evidence
The two captures Pulse comparedFull-page captures of Crusoe's pricing page (www.crusoe.ai) from Sep 21, 2026 and Sep 28, 2026; the change is marked around the middle of the page, where price reads $0.40 and then $2.50.
Sign in to see the evidence
The two full-page captures with this change marked up are free with an account.
Crusoe, the last 90 days
Full history- Oct 12, 2026packagingCrusoe adds DeepSeek V4.1 Flash to Serverless Inference at $0.29/1M input
- Oct 5, 2026packagingCrusoe drops Kimi K2.6 ($0.70/$3.50 per 1M) from Managed Inference lineup
- Sep 28, 2026increaseCrusoe: GPT-OSS 20B fine-tuning price $0.40->$2.50 + 2 more changes · this change
- Jul 27, 2026packagingCrusoe adds Nemotron 3 Ultra 550B to Serverless Inference at $1/$3.20 per…
- Jul 20, 2026improvedCrusoe: Launched Serverless Fine-Tuning + Self-Serve Deployments
Across Cloud & DevOps in the same window
All Cloud & DevOps changes- SpaceliftNov 2Spacelift rebuilds lineup: $399 Starter replaces quote-only, Free gets full card
- Tray.aiNov 2Tray.ai swaps "Talk to sales" for "Sized to your needs" on paid plans
- SaladNov 2Salad hikes RTX 5090 High-priority GPU rate 20% to $0.60/hr; Lowest tier flat
- ScalrNov 2Scalr spells out billable runs: failed runs count, drift detection is free
- RunPodNov 2RunPod raises Serverless GPU prices up to 25%, drops savings claim
Questions about this change
Did Crusoe raise prices in September 2026?
Yes. Pulse captured Crusoe's pricing page on Sep 21, 2026 and again on Sep 28, 2026 and recorded 1 price increase: $0.40 → $2.50 (+525%).
What changed on Crusoe's pricing page between Sep 21, 2026 and Sep 28, 2026?
Limit: 1M → 320K (−68%); feature: Not listed → 4 new Serverless Fine-Tuning models, added; price: $0.40 → $2.50 (+525%). Pulse compared the captures it took on Sep 21, 2026 and Sep 28, 2026; the before/after table above lists each item as written on the page.
How often does Crusoe change its pricing?
Pulse recorded 2 other pricing-page changes for Crusoe in the 90 days before this one; the previous change was "Crusoe adds Nemotron 3 Ultra 550B to Serverless Inference at $1/$3.20 per…" on Jul 27, 2026; Crusoe has been tracked since 2026-06-18.