Model release
Example page
The project, data and links on this page are made up, to show what dokk’s components can do. It does not document dokk itself.
Acme Large 2 is out today, with a bigger context window and better tool use.
Acme Large 2
Acme AI · acme-large-2
Our most capable model, for complex coding, analysis and long documents.
- Input
- $3 / 1M
- Output
- $15 / 1M
- Cached input
- $0.30 / 1M
- Context
- 200K
- Max output
- 64K
- Cutoff
- 2026-03
- Released
- 2026-10
Benchmarks #
- Acme Large 2
- 88.4%
- Acme Large 1
- 81.0%
- Acme Small 2
- 72.5%
- Acme Open 7B
- 48.2%
- Acme Large 2
- 420ms
- Acme Large 1
- 610ms
- Acme Small 2
- 180ms
Note
For latency, shorter bars are better. Our evals and their settings are in the technical report.
The family #
| Acme Large 2 | Acme Small 2 | Acme Open 7B | |
|---|---|---|---|
| Provider | Acme AI | Acme AI | Acme AI |
| API name | acme-large-2 |
acme-small-2 |
— |
| Input / 1M | $3 | $0.25 | — |
| Output / 1M | $15 | $1.25 | — |
| Context | 200K | 200K | 32K |
| Max output | 64K | 16K | — |
| Released | — | — | 2026-01 |
| License | — | — | Apache-2.0 |
| Features | text vision tools reasoning | text tools | text code |
Try it #
pip install acme-sdk
uv add acme-sdk
poetry add acme-sdk
from acme import Acme
client = Acme()
reply = client.chat(model="acme-large-2", message="Hi!")
print(reply.text)
System
Answer in one sentence.
What is new in Acme Large 2?
Acme Large 2
A 200K context window, better tool use and prompt caching, at the same price as before.
Open weights #
The small open model and its training data are on the Hugging Face Hub.
acme-ai/acme-open-7b
Hugging Face model
The open 7B model, for local use and fine-tuning.
acme-ai/acme-mix
Hugging Face dataset