AI Agent StoreAi2 · Byte-level research checkpoint
Blama 8B
A research model that reads and writes bytes instead of ordinary subword tokens. Ai2’s October 7 announcement highlights this new checkpoint; repository creation and modification dates are in August, so announcement and artifact dates stay separate.
Announced 2026-10-07 · First reviewed 2026-10-09 · Checked 2026-10-15. The announcement and discovery dates do not establish ordinary API availability. Unverified prices remain unknown.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on your own tasks before switching; results for another version or thinking setting may not apply.
What you can use today
Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact checkpoint ID: allenai/Llama-3-Blama-8B. This is a base autoregressive text checkpoint, rather than a documented tool-using chat assistant. Its card refers to custom Transformers code and counts max_new_tokens in bytes.
The card declares Llama 3 Community License. Blama’s card also uses the older Llama-B name and links; the current repository ID is kept exact. Stage 1 artifacts are training starting points, not extra completed base-model releases. Parent Qwen/Llama scores are not transferred.
The configuration’s max_position_embeddings is 8,192. That internal position allowance is not treated as a standard token-context limit. Byte budgets are not directly comparable to the token budgets of hosted chat models.
Hosted prices, output caps and reasoning controls for this exact checkpoint are unverified. Downloadable weights do not make inference free. Your bill can also include tools, retries and charged thinking tokens. Check the offer used by your application.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. A cheaper token price does not guarantee a cheaper completed task.
Compare caching and long-context costs ↗