The biggest recent price moves are concentrated in a few model families, with several DeepSeek variants jumping sharply and Beam cutting hard. For developers, that means budget volatility is real, especially if your stack leans on fast-moving hosted endpoints.
The largest increase in the set is openrouter/deepseek/deepseek-v4.1-flash (openrouter), up 386.1% from $0.11 to $0.52/Mtok blended. That is followed by openrouter/~deepseek/deepseek-pro-latest (openrouter), up 230% from $0.6 to $1.98/Mtok blended, and openrouter/~deepseek/deepseek-flash-latest (openrouter), up 150% from $0.21 to $0.52/Mtok blended. The pattern is clear, DeepSeek pricing moved up across multiple variants at once, which usually means the change is not isolated to a single SKU.
There is also a broader upward cluster outside DeepSeek. openrouter/~z-ai/glm-flash-latest (openrouter) doubled from $0.12 to $0.24/Mtok blended, while openrouter/qwen/qwen3-30b-a3b-instruct-2507 (openrouter) rose 77.8% from $0.08 to $0.15/Mtok blended. together_ai/Qwen/Qwen3.7-Max (together_ai) moved up 66.7% from $2.25 to $3.75/Mtok blended, which matters less in percentage terms than in absolute spend, because it is already the highest-priced item in this list.
On the downside, baseten/reflection/Beam-1.0 (baseten) fell 70.7% from $2.05 to $0.6/Mtok blended, and wandb/reflection/Beam-1.0-501B-A23B (wandb) fell 68.3% from $1.5 to $0.48/Mtok blended. That is the opposite of the DeepSeek pattern, and it shows why model routing can pay off, if your workload can tolerate switching providers or variants, the same class of model can become dramatically cheaper very quickly.
For budgeting, the practical takeaway is simple: recent price changes are large enough that static cost assumptions can go stale fast. If you are shipping against per-token margins, the difference between $0.11 and $0.52/Mtok, or between $2.05 and $0.6/Mtok, is big enough to change which model is viable for production.
| Model | Provider | Was | Now | Change |
|---|---|---|---|---|
| openrouter/deepseek/deepseek-v4.1-flash | Openrouter | $0.11 | $0.52 | +386.1% |
| openrouter/~deepseek/deepseek-pro-latest | Openrouter | $0.6 | $1.98 | +230% |
| openrouter/~deepseek/deepseek-flash-latest | Openrouter | $0.21 | $0.52 | +150% |
| openrouter/~z-ai/glm-flash-latest | Openrouter | $0.12 | $0.24 | +100% |
| openrouter/qwen/qwen3-30b-a3b-instruct-2507 | Openrouter | $0.08 | $0.15 | +77.8% |
| baseten/reflection/Beam-1.0 | Baseten | $2.05 | $0.6 | -70.7% |
| wandb/reflection/Beam-1.0-501B-A23B | Wandb | $1.5 | $0.48 | -68.3% |
| together_ai/Qwen/Qwen3.7-Max | Together_ai | $2.25 | $3.75 | +66.7% |
- DeepSeek saw the biggest upward moves, including 386.1% and 230% jumps.
- Beam got much cheaper, with drops of 70.7% and 68.3%.
- Qwen and GLM also moved up, so the price pressure is not limited to one family.
- Recheck model routing and budgets often, blended prices can shift fast enough to change your default choice.
About this data. Figures are from MyTokenTracker live data (model prices, AI Cost Index, open data API), free to use and cite under CC BY 4.0. Snapshot: November 9, 2026.