What changed in LLM hosting prices
The month's largest move: Fireworks AI cut Ultramarine 1 Flash in us by 33%. Prices moved at DeepInfra and Fireworks AI. 83 models are newly sold by a cloud. 116 were withdrawn. 46 retirements announced.
The last 30 days: 16 model launches, 9 price moves, 49 retirements set, 2 new places, 12 smaller changes, 114 retiring in 90 days.
Price cuts and rises, models and GPU instances added or withdrawn, new regions, retirements and the clouds' own announcements. Each item is found by comparing one collection of every cloud's price list with the one before it; none is written by hand.
The latest changes
- 20 November 2026: GLM 5.3 Flash: output price cut
- 20 November 2026: GLM 5.3 Flash: input price cut
- 20 November 2026: Ultramarine 1 Flash: output price cut
- 20 November 2026: Ultramarine 1 Flash: input price cut
- 18 November 2026: gemini-4-neon launched
- 19 November 2026: Telnyx now sells GPT OSS 120B (Preview) per token from $0.068 in / $0.30 out per 1M
- 20 November 2026: xAI now sells grok-build-0.1 per token from $1.00 in / $2.00 out per 1M
- 20 November 2026: GLM 5.3 Flash: cache reads cut
- 19 November 2026: Azure will retire DeepSeek-V4.1-Flash (2026-09-10) on 15 Dec 2026
- 19 November 2026: Azure will retire DeepSeek-V4-Pro-0813 (2026-08-13) on 15 Dec 2026
- 19 November 2026: Databricks now sells GPT-5.4 Nano per token from $0.20 in / $1.25 out per 1M in us-east-1
- 21 November 2026: Kimi K3.1 Pro: output price cut
- 21 November 2026: Kimi K3.1 Pro: input price cut
- 21 November 2026: Kimi K3.1 Pro: cache reads cut
- 20 November 2026: Together AI now sells tencent/Hy4 per token from $0.834 in / $2.50 out per 1M
- 20 November 2026: Together AI now sells Qwen/Qwen4-Max per token from $1.43 in / $4.28 out per 1M
- 20 November 2026: Fireworks AI now sells Kimi K3.1 Pro per token from $1.00 in / $5.00 out per 1M
- 20 November 2026: Fireworks AI now sells DeepSeek V4.1 Flash 1017 Ultra-fast (Preview) per token from $0.60 in / $2.40 out per 1M
- 20 November 2026: Fireworks AI now sells DeepSeek V4.1 Flash 1017 Flex per token from $0.075 in / $0.30 out per 1M
- 20 November 2026: Ultramarine 1 Flash: cache reads cut
- 20 November 2026: DeepInfra now sells RadixArk/DaisyRose-27B per token from $0.12 in / $1.50 out per 1M
- 20 November 2026: Databricks now sells Long context per token from $5.00 in / $15.00 out per 1M in us-east-1
- 19 November 2026: Telnyx now sells Kimi K3.1 (Preview) per token from $2.00 in / $10.00 out per 1M
- 19 November 2026: AWS now sells Stability AI Image Services per token in 3 regions
- 19 November 2026: Google Cloud: Feature: Gemini 4 Hydrogen in CodeMender (Preview)
- 19 November 2026: Azure will retire MAI-Transcribe-2-Streaming (2026-08-06) on 4 Dec 2026
- 19 November 2026: Azure moved the retirement of MAI-Image-2.5 (2026-06-02) from 1 Oct 2026 to 24 Nov 2026
- 19 November 2026: Azure moved the retirement of MAI-Image-2.5-Flash (2026-06-02) from 1 Oct 2026 to 24 Nov 2026
- 19 November 2026: Azure moved the retirement of MAI-Image-2.5-Pro (2026-06-19) from 1 Oct 2026 to 24 Nov 2026
- 19 November 2026: Databricks now sells GPT-6.2 Sol per token from $2.00 in / $10.00 out per 1M in us-east-1
- 19 November 2026: Databricks now sells GPT-6.2 Astra per token from $8.00 in / $40.00 out per 1M in us-east-1
- 19 November 2026: Azure moved the retirement of claude-fable-5 (1) from 5 Dec 2027 to 7 Dec 2027
- 19 November 2026: Azure moved the retirement of claude-fable-5-1 (1) from 5 Dec 2027 to 7 Dec 2027
- 19 November 2026: Azure will retire grok-4.7 (1) on 5 Oct 2027
- 19 November 2026: Azure will retire grok-4.6 (1) on 24 Aug 2027
- 19 November 2026: Azure will retire grok-4.3 (1) on 12 May 2027
- 19 November 2026: Azure moved the retirement of DeepSeek-V4-Pro (2026-04-23) from 20 Feb 2028 to 22 Feb 2028
- 19 November 2026: Azure moved the retirement of DeepSeek-V4-Flash (2026-04-23) from 20 Feb 2028 to 22 Feb 2028
- 19 November 2026: Azure moved the retirement of DeepSeek-V3.2 (1) from 1 Nov 2027 to 2 Nov 2027
- 19 November 2026: Azure will retire gpt-6.1-sol (2026-09-29) on 11 Mar 2028