Skip to content
Navigation Menu
Sign in
Appearance settings
Platform
AI CODE CREATION
GitHub Copilot
Write better code with AI
GitHub Copilot app
Direct agents from issue to merge
MCP Registry
Integrate external tools
DEVELOPER WORKFLOWS
Actions
Automate any workflow
Codespaces
Instant dev environments
Issues
Plan and track work
Code Review
Manage code changes
Code Quality
Enforce quality at merge
APPLICATION SECURITY
GitHub Advanced Security
Find and fix vulnerabilities
Code security
Secure your code as you build
Secret protection
Stop leaks before they start
EXPLORE
Why GitHub
Documentation
Blog
Changelog
Marketplace
View all features
Solutions
BY COMPANY SIZE
Enterprises
Small and medium teams
Startups
Nonprofits
BY USE CASE
App Modernization
DevSecOps
DevOps
CI/CD
View all use cases
BY INDUSTRY
Healthcare
Financial services
Manufacturing
Government
View all industries
View all solutions
Resources
EXPLORE BY TOPIC
AI
Software Development
DevOps
Security
View all topics
EXPLORE BY TYPE
Customer stories
Events & webinars
Ebooks & reports
Business insights
GitHub Skills
SUPPORT & SERVICES
Documentation
Customer support
Community forum
Trust center
Partners
View all resources
Open Source
COMMUNITY
GitHub Sponsors
Fund open source developers
PROGRAMS
Security Lab
Maintainer Community
GitHub Stars
Archive Program
REPOSITORIES
Topics
Trending
Collections
Enterprise
ENTERPRISE SOLUTIONS
Enterprise platform
AI-powered developer platform
AVAILABLE ADD-ONS
GitHub Advanced Security
Enterprise-grade security features
Copilot for Business
Enterprise-grade AI features
Premium Support
Enterprise-grade 24/7 support
Pricing
Search
/
Sign in
Sign up
Appearance settings
You signed in with another tab or window.
Reload
to refresh your session.
You signed out in another tab or window.
Reload
to refresh your session.
You switched accounts on another tab or window.
Reload
to refresh your session.
Dismiss alert
{{ message }}
truespar
/
paddock
Public
Notifications
You must be signed in to change notification settings
Fork
17
Star
115
Code
Issues
7
Pull requests
0
Actions
Projects
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Actions
Projects
Security and quality
Insights
Commits
Branch selector
main
User selector
All users
Datepicker
All time
Commit history
Commits on Sep 29, 2026
Fixes a Qwen 3.8 server on the DGX Spark holding memory it never used, which could run it out of memory after about a day, and requests with many images or long audio clips now use far less memory.…
Show description for 5bc1453
panterlo
committed
5bc1453
View commit details
Copy full SHA for 5bc1453
Browse repository at this point
Commits on Sep 28, 2026
Version 0.1.11
panterlo
committed
f99cacd
View commit details
Copy full SHA for f99cacd
Browse repository at this point
Flash-Next fixes: prefill no longer reads stale data from reused cache pages, and a reply generated while speculating keeps its cache checkpoints, so the next turn resumes from it.
panterlo
committed
4ea5867
View commit details
Copy full SHA for 4ea5867
Browse repository at this point
Resume checkpoints for Qwen 3.5-3.8, Nemotron and Qwen 3.8 Flash-Next now live in the KV cache instead of a fixed reserve, so more context fits on the same GPU. Flash-Next's KV cache is now paged, …
Show description for 023039e
panterlo
committed
023039e
View commit details
Copy full SHA for 023039e
Browse repository at this point
Commits on Sep 25, 2026
Version 0.1.10: Laya, an open decision model for the Reads page, image input and conditional questions in reads, and faster Qwen 3.8 Flash-Next
panterlo
committed
610cd8f
View commit details
Copy full SHA for 610cd8f
Browse repository at this point
New model: Laya, an open decision model that answers typed questions about a text with calibrated probabilities, downloadable from the Studio and used by the Reads page. Reads take images, several …
Show description for 39d5dba
panterlo
committed
39d5dba
View commit details
Copy full SHA for 39d5dba
Browse repository at this point
Move Studio state and Reads history to shared SQLite
Show description for 35952eb
panterlo
committed
35952eb
View commit details
Copy full SHA for 35952eb
Browse repository at this point
Fix native Reads workflows, persistent history, and notification previews
Show description for 5b86727
panterlo
committed
5b86727
View commit details
Copy full SHA for 5b86727
Browse repository at this point
Commits on Sep 24, 2026
Check the three-row Studio footer at all sidebar widths
panterlo
committed
d4b4548
View commit details
Copy full SHA for d4b4548
Browse repository at this point
Use Paddock's corrected mark for the macOS 0.1.9 release
Show description for ee7ae34
panterlo
committed
ee7ae34
View commit details
Copy full SHA for ee7ae34
Browse repository at this point
Version 0.1.9: Whisper endpoints can load their model on demand and unload it when idle, Qwen 3.8 Flash-Next uses its own sparse attention on long prompts and loads its full context window, long co…
Show description for fd34f33
panterlo
committed
fd34f33
View commit details
Copy full SHA for fd34f33
Browse repository at this point
New models: DiffusionGemma 26B A4B (Google's diffusion LLM) with a Jev-style structured-read API and a Reads page in the Studio, Qwen-Image 2.1 text-to-image and image editing with an image lane in…
Show description for 845eb5b
panterlo
committed
845eb5b
View commit details
Copy full SHA for 845eb5b
Browse repository at this point
Commits on Sep 21, 2026
Clarify macOS preview status and download options
panterlo
committed
249cf25
View commit details
Copy full SHA for 249cf25
Browse repository at this point
Fix Metal model restarts with warm file caches
panterlo
committed
93134ff
View commit details
Copy full SHA for 93134ff
Browse repository at this point
Version 0.1.8
panterlo
committed
1b321df
View commit details
Copy full SHA for 1b321df
Browse repository at this point
macOS: Metal model management fixes, shared GPU telemetry, and release packaging, signing and notarization tooling
panterlo
committed
970ec92
View commit details
Copy full SHA for 970ec92
Browse repository at this point
Native macOS app and Metal backend for Apple Silicon (builds from source, not yet a release), Bonsai 2 27B with image input, Mac model packages for Qwen 3.8, Gemma 4, Muse and Flash-Next, Qwen 3.8 …
Show description for d672bd7
panterlo
committed
d672bd7
View commit details
Copy full SHA for d672bd7
Browse repository at this point
Commits on Sep 20, 2026
Merge pull request #31 from DivyamTalwar/fix/studio-summary-budget
panterlo
committed
a08538b
View commit details
Copy full SHA for a08538b
Browse repository at this point
Merge pull request #32 from DivyamTalwar/fix/studio-timeout
panterlo
committed
fc1cfe1
View commit details
Copy full SHA for fc1cfe1
Browse repository at this point
Merge pull request #30 from DivyamTalwar/fix/studio-model-race
panterlo
committed
1d33dbb
View commit details
Copy full SHA for 1d33dbb
Browse repository at this point
Merge pull request #28 from DivyamTalwar/fix/studio-input-budget
panterlo
committed
115cdf1
View commit details
Copy full SHA for 115cdf1
Browse repository at this point
Commits on Sep 18, 2026
studio: charge injected summaries to the reply budget
Show description for 4bca56c
DivyamTalwar
committed
4bca56c
View commit details
Copy full SHA for 4bca56c
Browse repository at this point
studio: bound background compaction requests and retries
Show description for 051b760
DivyamTalwar
committed
051b760
View commit details
Copy full SHA for 051b760
Browse repository at this point
studio: discard compaction results after a model switch
Show description for 84374b0
DivyamTalwar
committed
84374b0
View commit details
Copy full SHA for 84374b0
Browse repository at this point
studio: skip compaction when no transcript budget remains
Show description for 40e6e1c
DivyamTalwar
committed
40e6e1c
View commit details
Copy full SHA for 40e6e1c
Browse repository at this point
Commits on Sep 17, 2026
Tool calls work on Qwen 3.8 Flash Next, and a stopped model no longer shows as running forever
panterlo
committed
c8d97fb
View commit details
Copy full SHA for c8d97fb
Browse repository at this point
Commits on Sep 16, 2026
Version 0.1.7
panterlo
committed
98e6703
View commit details
Copy full SHA for 98e6703
Browse repository at this point
Qwen 3.8 Flash Next performance optimizations and MTP support
panterlo
committed
d1bad47
View commit details
Copy full SHA for d1bad47
Browse repository at this point
Commits on Sep 13, 2026
Version 0.1.6: the DGX Spark is supported
panterlo
committed
7370581
View commit details
Copy full SHA for 7370581
Browse repository at this point
Gemma 4 reads small print and keeps to its memory budget, conversations resume from the last reply across Nemotron, Qwen and Flash-Next, Nemotron and 4-bit models much faster on the DGX Spark, and …
Show description for 3a395a6
panterlo
committed
3a395a6
View commit details
Copy full SHA for 3a395a6
Browse repository at this point
Commits on Sep 8, 2026
DGX Spark support and arm64 builds, Qwen 3.8 Flash-Next faster and in the catalog as a small file, quicker Qwen 3.5 prompts on small cards, Granite tasks run page by page
panterlo
committed
7d0c62a
View commit details
Copy full SHA for 7d0c62a
Browse repository at this point
Commits on Sep 7, 2026
Version 0.1.5
panterlo
committed
ba54467
View commit details
Copy full SHA for ba54467
Browse repository at this point
Prefill on 1-3 bit files runs on the tile GEMM, level with the 4-bit files
panterlo
committed
18168fa
View commit details
Copy full SHA for 18168fa
Browse repository at this point
Merge pull request #21 from NodeNestor/iquant-tile
panterlo
committed
ad23041
View commit details
Copy full SHA for ad23041
Browse repository at this point
MoE expert offload: faster prefill waves; laguna and nemotron: batched serving, scratch sized on demand
panterlo
committed
ff52aad
View commit details
Copy full SHA for ff52aad
Browse repository at this point
Previous
Next
You can’t perform that action at this time.
