AI API & GPU price report: October 2026
In October 2026 we recorded 309 price changes across AI model APIs and GPU clouds: 160 API price drops, 95 increases and 47 GPU rental price changes. The biggest cut was GLM 5.3 Flash on OpenRouter (−86%).
GPU price index
On-demand price per GPU-hour across the GPU clouds we track, at the end of October 2026, compared with the start of the month.
| GPU | Lowest | Cheapest on | Median | Median vs start | Offers |
|---|---|---|---|---|---|
| B300 | $7.10 | Modal | $9.14 | +23% | 12 |
| B200 | $3.50 | Vultr | $8.19 | +34% | 17 |
| H200 | $2.63 | Vast.ai | $5.12 | +23% | 20 |
| H100 | $1.68 | Latitude.sh | $3.50 | +7% | 38 |
| A100 80GB | $0.31 | Vast.ai | $1.80 | +14% | 31 |
| L40S | $0.47 | Vast.ai | $1.56 | +4% | 14 |
| RTX PRO 6000 | $0.92 | Vast.ai | $2.45 | +30% | 17 |
| RTX 5090 | $0.28 | Vast.ai | $0.74 | +5% | 4 |
| RTX 4090 | $0.27 | Vast.ai | $0.64 | +27% | 7 |
| MI300X | $1.85 | Vultr | $3.45 | +30% | 5 |
Biggest API price drops
Price drops(160)
- GLM 5.3 FlashonOpenRouterinput $0.15 → $0.0513/1M, output $0.5 → $0.0701/1M↓−86%
- Mistral Large 3onAmazon Bedrockinput $3 → $0.5/1M, output $9 → $1.5/1M↓−83%
- GLM 5.2onOpenRouterinput $0.38 → $0.189/1M, output $3.49 → $0.594/1M↓−83%
- DeepSeek V4 Pro (0813)onOpenRouterinput $0.17 → $0.132/1M, output $1.94 → $0.396/1M↓−80%
- GPT-5.4onWaveSpeed AIinput $10 → $2.5/1M, output $30 → $15/1M↓−75%
- DeepSeek V4 ProonParasailinput $1.74 → $0.45/1M↓−74%
- Kimi K3onOpenRouterinput $0.72 → $0.1919/1M, output $13 → $7.697/1M↓−73%
- InklingonVenice AIinput $1.25 → $0.5625/1M, output $5.063 → $1.5/1M↓−70%
- Qwen3.6 35B A3BonOpenRouterinput $0.15 → $0.05/1M, output $1 → $0.7/1M↓−67%
- Claude Opus 4.6onWaveSpeed AIinput $15 → $5/1M, output $75 → $25/1M↓−67%
…and 150 more; see the daily pages under price changes.
Biggest API price increases
Price increases(95)
- Llama 3.3 70BonWaveSpeed AIinput $0.1 → $0.72/1M, output $0.32 → $0.72/1M↑+620%
- DeepSeek V4 FlashonOpenRouterinput $0.09 → $0.0224/1M, output $0.18 → $1.28/1M↑+611%
- Veo 3.1onKIE AIper request $0.025 → $0.15/request↑+500%
- DeepSeek V4 Flash (0731)onWaveSpeed AIinput $0.06 → $0.22/1M, output $0.12 → $0.66/1M↑+450%
- DeepSeek V3.1onAIMLAPIinput $0.294 → $0.894/1M, output $0.441 → $2.338/1M↑+430%
- Qwen 3.5 35BonAlibaba Cloudinput $0.1 → $0.25/1M, output $0.4 → $2/1M↑+400%
- PixVerse V6onfal.aiper second $0.005 → $0.025/sec↑+400%
- DeepSeek V4 ProonDeepSeekinput $0.435 → $1.32/1M, output $0.87 → $3.96/1M↑+355%
- Llama 3.1 8B InstructonWaveSpeed AIinput $0.05 → $0.22/1M, output $0.08 → $0.22/1M↑+340%
- DeepSeek V4.1 FlashonDeepSeekinput $0.14 → $0.3/1M, output $0.28 → $1.2/1M↑+329%
GPU rental price changes
GPU rental prices(47)
- RTX 5090onVast.aion-demand $0.43 → $0.28/hr↓−35%
- L40onVast.aion-demand $0.45 → $0.33/hr↓−27%
- RTX 4090onVast.aion-demand $0.35 → $0.27/hr↓−23%
- L4onVast.aion-demand $0.32 → $0.25/hr↓−22%
- RTX PRO 6000 (Server)onVast.aion-demand $1.33 → $1.06/hr↓−20%
- H200onVast.aion-demand $3.29 → $2.63/hr↓−20%
- RTX 3090onVast.aion-demand $0.13 → $0.11/hr↓−15%
- RTX PRO 6000onRuncrateon-demand $2.85 → $2.45/hr↓−14%
- RTX 4090onTensorDockon-demand $0.37 → $0.35/hr↓−5%
- AMD MI325XonVultrreserved $2.1 → $2/hr↓−5%
Most active providers
- OpenRouter123 changes
- Venice AI40 changes
- WaveSpeed AI31 changes
- Verda12 changes
- AIMLAPI10 changes
- RunPod10 changes
- Vast.ai10 changes
- fal.ai9 changes
- KIE AI7 changes
- Cloudflare Workers AI6 changes
Using these numbers
You're welcome to quote this report. Please link to it as the source:
Source: Inference Hub, AI API & GPU Price Report, October 2026 (https://inferencehub.org/price-report/2026-10)
How we collect prices: from each provider's own pricing page or API, checked weekly; see our methodology. Percentages compare the new price with the old; when input and output prices both moved, the larger move is shown. 7 changes of 10× or more are left out of the lists above, since they're almost always unit or listing corrections rather than price moves. Get daily alerts by RSS or on Telegram.