gpudiff the public record of change in GPU cloud pricing

Follow the diffs: RSS · Free API · Sponsor this site

gpt-oss-20b API price — 5 gateways compared

Cheapest right now: OpenRouter at $0.02/MTok in — 3.9× cheaper than the priciest gateway.

from $0.02 per million input tokens · 3.9× spread between gateways · updated continually

Gateway	$ in /MTok	$ out /MTok	Context	Model ID
OpenRouter fees	$0.02	$0.09	131,072	openai/gpt-oss-20b	use
Glama	$0.03	$0.13	131,072	openai/gpt-oss-20b	use
DeepInfra	$0.03	$0.14	131,072	openai/gpt-oss-20b	use
Novita	$0.04	$0.15	131,072	openai/gpt-oss-20b	use
Requesty	$0.07	$0.30	131,072	fireworks/gpt-oss-20b	use

Fees beyond the token price

OpenRouter: No markup on token prices (provider pass-through), but credit purchases cost 5.5% via card ($0.80 minimum) or 5% via crypto; BYOK usage above the monthly allowance is charged 5%. source

2026-09-22 openrouter cut openai-gpt-oss-20b-output 30.8%: 0.13 → 0.09 usd_per_mtok
2026-09-22 openrouter cut openai-gpt-oss-20b-input 40.0%: 0.03 → 0.018 usd_per_mtok
2026-09-11 novita listed gpt-oss-20b-output at 0.15 usd_per_mtok
2026-09-11 novita listed gpt-oss-20b-input at 0.04 usd_per_mtok
2026-09-10 novita delisted gpt-oss-20b-output
2026-09-10 novita delisted gpt-oss-20b-input
2026-08-27 requesty raised gpt-oss-20b-output 11.1%: 0.27 → 0.3 usd_per_mtok
2026-08-27 requesty raised gpt-oss-20b-input 11.1%: 0.063 → 0.07 usd_per_mtok

The cheapest tracked gateway charges $0.02 per million input tokens and $0.09 per million output tokens (OpenRouter). Prices refresh continually.

OpenRouter for input tokens and OpenRouter for output tokens, among the 5 gateways we track.

Yes — the most expensive tracked gateway charges 3.9× the cheapest for input tokens. Identical model, identical tokens, different bill.