Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
Commits on main
sample: define multi-row distributions without a draft chain
0fcfc99 Jesse Gross committed 3 months ago
mlxrunner: let each model declare the cache slots it needs
e7fbd52 Jesse Gross committed 3 months ago
mlxrunner: return the draft-conditioning state from a model forward
2f84872 Jesse Gross committed 3 months ago
openai: expand namespace tool declarations in the responses API (#17593)
35f7138 Daniel Hiltgen committed 3 months ago
mlxrunner: stop cache rewind refills from corrupting later lazy snapshots
1448938 Jesse Gross committed 3 months ago
progress: fix data races on ticker, states, spinner, and bar state (#17445)
43983ed Daniel Hiltgen committed 3 months ago
openai: match openai's streaming wire format for chat completions (#17485)
8edecb5 Bruce MacDonald committed 3 months ago
Merge pull request #17483 from ollama/drifkin/suggest-cloud
8d8c701 Devon Rifkin committed 3 months ago
app/updater: drain background update-check goroutine before returning (#17446)
b63eed9 Daniel Hiltgen committed 3 months ago
qwen3_5: load and run the MTP head as a speculative draft
4f9d09e Jesse Gross committed 5 months ago
qwen3_5: pack GDN input projections into one layout at load
bd3f22e Jesse Gross committed 4 months ago
mlxrunner: count every speculative round and log stats at info
3ff2dcb Jesse Gross committed 3 months ago