NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
Go modules, #1 by repository stars
Latest stable
v0.40.17
Last release today
21 Nov 2026
Ships on a steady schedule
a new release about every 8 days
Some releases are documented
notes for 32 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
3 years old
1371 releases · first in 2024
Drag across the ruler to compare, or tap a release to open it.
pplx-decider is a 27B decision model from Perplexity, fine-tuned from Qwen3.8-27B
ollama show now reports which renderer and parser a model uses and where its chat template comes from; the APIrenderer, parser, and template_source fieldsollama launch claude now clamps CLAUDE_CODE_AUTO_COMPACT_WINDOW to Claude Code's supported range, avoiding aFull Changelog: v0.40.6...v0.40.7
Deduplicate MLX CUDA runtime payloads
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Move CUDA libraries shared by the GGML and MLX backends into the sibling CUDA payload, and teach in-process and subprocess MLX loading to resolve that packaged dependency directory.
Keep MLX-private libraries and JIT headers in the MLX overlay, include the full OpenBLAS Fortran runtime closure, and verify those dependencies in release packaging.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Implement the Qwen3.5-backed decision scorer with noncausal attention, the publisher readout, image handling, and System One prompt encoding.
Implement the Qwen3.5-backed decision scorer with noncausal attention,
the publisher readout, image handling, and System One prompt encoding.
Nothing published for this version
Nothing published for this version
win: prevent orphaned runners by @dhiltgen in #18500
Full Changelog: v0.40.5...v0.40.6-rc0
Nothing published for this version
Behind a proxy, defer redirect-target DNS to the proxy; without one, keep the resolve-and-pin path. Trust default-registry/CDN hostnames with no local
Behind a proxy, defer redirect-target DNS to the proxy; without one, keep the resolve-and-pin path. Trust default-registry/CDN hostnames with no local DNS. Fix a nil CancelFunc crash in blobDownload.release.
Nothing published for this version
Nothing published for this version
NVIDIA Nemotron 3.5 Super is a vision-language model for coding, reasoning, tool use, and document understanding. It pairs a 120B-parameter hybrid Mam
NVIDIA Nemotron 3.5 Super is a vision-language model for coding, reasoning, tool use, and document understanding. It pairs a 120B-parameter hybrid Mamba-2/Transformer mixture-of-experts backbone (12B active per token) with a C-RADIO vision encoder, and supports tool calling, an on/off thinking mode, and context up to 1M tokens.
ollama run nemotron-3.5-super
A compact sibling, Nemotron 3.5 Bolt, distills Super down to 74B parameters (9B active) tuned to fit a 24 GB NVIDIA GPU for everyday agentic tasks, with speculative decoding via its included drafter.
clef:cloud and clef-flash:cloud to run decision models through ollama.com cloud without pulling them locallyFull Changelog: v0.40.4...v0.40.5
llama.cpp: version bump b11669
llama.cpp: version bump b11669 (#19032)
ci: shorten Windows Vulkan build path
Keep the nested shader-generator compiler probes below Windows path limits in checkouts with longer repository names.
Accept published Nemotron layer names while retaining legacy aliases, and
fix layer-count inference and detached MTP quantization. Stabilize Mamba
softplus, preserve FP32 vision normalization, and compile Mamba gate and
norm scaling.
Support per-expert F32 global scales and preserve activation-scale
companions in ModelOpt and compressed-tensors NVFP4 imports. Normalize
reciprocal globals and reject invalid scales. Keep activation scales
separate from weight-only inference.
mlx: add Nemotron 3.5 Super
minor adjustments
Nothing published for this version
Nothing published for this version
Nothing published for this version
MLX: version bump
MLX: version bump (#18993)
Nothing published for this version
Nothing published for this version
Nothing published for this version
Decision models can now be used directly from the terminal with ollama run — ask yes/no questions, classify text into labels, or score against a rubri
Decision models can now be used directly from the terminal with ollama run — ask yes/no questions, classify text into labels, or score against a rubric:
ollama run nimble --ask "Is a refund being requested?" < ticket.txt
ollama launch can now be used with usage credits without upgrading to a paid planollama launch codex now includes locally installed Ollama models in its model picker. Use /model within the Codex CLI to swap models.Full Changelog: v0.40.3...v0.40.4
Nothing published for this version
launch: allow usage credits without upgrading
launch: allow usage credits without upgrading (#18985)
Add Kev decision scoring to MLX
Load Kev models with their FP32 decision head and release metadata.
Support System One scoring with shared-state caching and Kev-specific limits.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
MLX, MLX-C: version bump
MLX, MLX-C: version bump (#18956)
Nothing published for this version
ollama launch claude no longer shows a warning about Claude connectors by @hoyyeva in #18878
ollama launch claude no longer shows a warning about Claude connectors by @hoyyeva in #18878ollama launch codex no longer warns about service tiers by @hoyyeva in #18879Full Changelog: v0.40.2...v0.40.3
server: skip local model compatibility migration
server: skip local model compatibility migration (#18908)
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Models downloaded with earlier versions of Ollama are upgraded in the background the first time you run them, for better performance and compatibility
Models downloaded with earlier versions of Ollama are upgraded in the background the first time you run them, for better performance and compatibility when running on llama.cpp.
To make downgrading safe, Ollama keeps the original copy as a backup, so upgraded models are temporarily kept on disk. A future release will remove these backups automatically.
To remove backed up models now (requires jq):
curl -s localhost:11434/api/tags | jq -r '.models[].name' | while read -r m; do
curl -s localhost:11434/api/show -d "{\"model\":\"$m\"}" | jq -r \
'select(any(.manifests[]?; .runner == "llamacpp")) | .manifests[] | select(.runner == "ggml") | .digest'
done | sort -u | xargs -n1 ollama rmThis only deletes the backups. If you later downgrade to a version older than 0.40, you'll need to re-pull those models.
Other changes
ollama list no longer shows duplicate entries for upgraded models by @dhiltgen in #18874ollama launch claude uses the model's full context length by @jberg5 in #18855Full Changelog: v0.40.1...v0.40.2-rc0
When a legacy GGUF which requires our llama.cpp patch set is loaded, we do a lazy conversion on disk to make it llama.cpp compatible. To support downg
When a legacy GGUF which requires our llama.cpp patch set is loaded, we do a
lazy conversion on disk to make it llama.cpp compatible. To support
downgrades, we temporarily hold both the new and old GGUFs and create a shadow
v1 manifest tag to prevent the downgraded server from deleting the new blobs.
Showing these in the list output is confusing to users and API consumers. The
legacy copies will be removed in a future release to recover the disk space.
Nothing published for this version
server: proxy cloud usage and balance APIs by @drifkin in #18829
Full Changelog: v0.40.0...v0.40.1-rc0
Nothing published for this version
mlx: drop carried metal residency patch now that it is upstream
mlx: drop carried metal residency patch now that it is upstream (#18854)
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Models run on MLX on Apple Silicon by default
Models run on MLX on Apple Silicon by default
In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX.
ollama pull qwen3.8
ollama run qwen3.8
Additional models include gemma4, qwen3.6 and qwen3.5
Decision models are now available on MLX as well: Nimble tev1 clef clef-flash
MLX now has support for an embedding model: embeddinggemma-2
We will continue testing and enabling additional models.
Full Changelog: v0.35.1...v0.40.0
Implements the EmbeddingGemma2Model architecture on the MLX runner: 24-layer bidirectional text encoder with PLE, shared gemma4 vision/audio towers, m
Implements the EmbeddingGemma2Model architecture on the MLX runner: 24-layer
bidirectional text encoder with PLE, shared gemma4 vision/audio towers,
mean-pool + L2 output. /api/embed accepts per-item media via input dicts.
mlx: fix patch for recent mlx update
mlx: fix patch for recent mlx update (#18812)
add scopes for unit tests to reduce memory usage
MLX: version bump
add scopes for unit tests to reduce memory usage
Strip off pre-release from the version string so RCs can pull models for the same release.
Strip off pre-release from the version string so RCs can pull
models for the same release.
Merge remote-tracking branch 'upstream/main' into release_v0.40.0
Merge remote-tracking branch 'upstream/main' into release_v0.40.0
mlx: match publisher tokenizer semantics
Honor pretokenizer stage order, split behavior, Unicode boundaries,
added-token normalization, and ranked BPE merges. Handle empty added
tokens and empty Metaspace input consistently.
Add shared Go/Python reference cases using published tokenizers, pulling
missing models directly and failing on errors, plus focused regressions
for configuration precedence, byte fallback, and parallel encoding.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Ollama now supports Clef and Clef Flash , Cloudflare's new open-source decision models, through /v1/systemone .
Ollama now supports Clef and Clef Flash, Cloudflare's new open-source decision models, through /v1/systemone.
Clef (27B) and Clef Flash (9B) are multimodal: requests can now include images alongside the text state, shared by all questions and scored jointly with it.
curl http://localhost:11434/v1/systemone -d '{
"model": "clef-flash",
"state": "The user took this screenshot.",
"images": ["<base64-encoded image>"],
"questions": {
"has_ollama": {"type": "noul", "instructions": "Does this image contain Ollama?"}
}
}'{
"model": "clef-flash",
"answers": {
"has_ollama": {
"type": "noul",
"noul": 0.958
}
},
"usage": {
"input_tokens": 548,
"output_tokens": 0
}
}
CAPABILITY declarations, so model creators can explicitly declare what a model can do. Declarations are preserved when creating from GGUF or safetensors, through model inheritance, and on Modelfile exportollama show and the model list now report only decision as the capability for decision models, so clients no longer offer them for general chat, tools, or thinkingFull Changelog: v0.35.0...v0.35.1
One column per month.