OpenScale

The cheapest and fastest way to run open models.
Intelligence too cheap to meter.

[OpenScale @openscale_hq]

OpenScale is the cheapest way to run open-source models.

We're starting with DeepSeek V4.1 Flash at $3 per billion cached tokens ($0.003 per 1M). You read that right: billion, not million.

It's our first step toward intelligence too cheap to meter.

[OpenScale @openscale_hq]

Get $5 free to start: https://t.co/vT1TZdjLlh

[OpenScale @openscale_hq]

GLM 5.3 Flash is live on OpenScale.

Cheapest anywhere: $0.12 in, $0.024 cached, $0.40 out per 1M tokens, 20% below https://t.co/RKqlrfXNe3's own API.

Frontier-level coding and agent performance, 20x cheaper than Claude Opus 5.5.

[OpenScale @openscale_hq]

Get $5 free and try it: https://t.co/vT1TZdjLlh

[OpenScale @openscale_hq]

DeepSeek V4.1 Flash-1017 just dropped.
The cheapest, fastest place to run it is OpenScale.

Smarter: 36% higher on Terminal-Bench 4.0
Faster: 200+ tok/s, 16% fewer reasoning tokens
Cheapest: $0.06 in, $0.003 cached, $0.24 out per 1M

[OpenScale @openscale_hq]

Try it with $5 in free credits when you sign up: https://t.co/P6KLTAm1Og

More open-source models are coming next. Stay tuned ⚡️

[OpenScale @openscale_hq]

Agentic workloads are ~99% input tokens, and almost all of that can be cached.

On that mix, OpenScale has by far the lowest effective rate for DeepSeek V4.1 Flash anywhere: about $7 per billion tokens, 43% cheaper than the next provider.

[OpenScale @openscale_hq]

OpenScale is the cheapest way to run open-source models.

We're starting with DeepSeek V4.1 Flash at $3 per billion cached tokens ($0.003 per 1M). You read that right: billion, not million.

It's our first step toward intelligence too cheap to meter.