gpudiff the public record of change in GPU cloud pricing
Follow the diffs: RSS · Free API · Sponsor this site

Where to buy LLM tokens: the cheapest gateway right now

One pick per tier — the lowest live per-token price across six gateways, refreshed hourly, with a volatility read. Same model, wrong gateway, and you can pay several times as much.

Cheapest frontier model (Claude / GPT / Gemini class)

Claude Sonnet 5 — $1.80 in / $9.00 out per MTok
cheapest via Requesty new · use → · all gateways & history
Lowest input price across the gateways we track for this tier. Frontier models are usually identical across gateways (list passthrough); open-weight prices move, so watch the volatility tag.

Cheapest strong open-weight model (Llama / Qwen / DeepSeek / GLM)

Llama 4 Maverick — $0.18 in / $0.77 out per MTok
cheapest via Requesty new · use → · all gateways & history
Lowest input price across the gateways we track for this tier. Frontier models are usually identical across gateways (list passthrough); open-weight prices move, so watch the volatility tag.

Cheapest fast/cheap workhorse (flash-class)

DeepSeek V4 Flash — $0.08 in / $0.16 out per MTok
cheapest via OpenRouter volatile · use → · all gateways & history
Lowest input price across the gateways we track for this tier. Frontier models are usually identical across gateways (list passthrough); open-weight prices move, so watch the volatility tag.

Full table, every model, and per-model history on the LLM prices page. These are published list prices, not quotes; see methodology.