Bridge1 resources

https://bridge1.ai/resources/Bridge1 resources2026-10-30T00:00:00Zhttps://bridge1.ai/resources/articles/local-model-image-input/Local coding model vision: will it receive my screenshot?2026-10-29T00:00:00Z2026-10-30T00:00:00ZBridge1

For Mac developers sending screenshots to a local model, check the exact profile’s vision label and retain the image settings supplied for your client. This identifies a supported image-input path; a useful text description or an attachment control alone does not prove the model received an image.

https://bridge1.ai/resources/articles/local-model-memory-context-mac/Why model download size is not its Mac memory requirement2026-10-29T00:00:00Z2026-10-30T00:00:00ZBridge1

Size a local coding session on Mac around the exact model profile, the source and history it must carry, and the apps you need open. Download storage, running memory and context capacity answer separate questions; keep measured depth beside any estimate for a larger window.

https://bridge1.ai/resources/articles/mlx-gguf-quantization-labels/Read MLX and GGUF labels before choosing a model2026-10-29T00:00:00Z2026-10-30T00:00:00ZBridge1

For Mac developers comparing local coding models, choose the exact catalog profile before interpreting its MLX, GGUF or precision label. Those labels help identify what you will run, but they do not establish equal quality, speed, image support or memory use across profiles.

https://bridge1.ai/resources/articles/pi-dev-opencode-beginners-first-project/Pi.dev and OpenCode for beginners: your first project2026-10-30T00:00:00Z2026-10-30T00:00:00ZBridge1

Ask a local model to write a page for a made-up reading club, then check the saved file yourself. Use OpenCode through Bridge1, with a practice folder for the page and a few notes to summarize.

https://bridge1.ai/resources/articles/read-local-model-benchmarks/SWE-bench and local quants: how to read model results2026-10-29T00:00:00Z2026-10-30T00:00:00ZBridge1

A SWE-bench score for base weights does not establish the coding quality of a local quant. For Mac developers choosing a local model, compare the tested artifact and quantization before using that score, and keep quality results separate from runtime speed. Neither establishes your task result on another Mac without relevant evidence.

https://bridge1.ai/resources/models/ornith-1-5-35b-a3b-mac/Ornith 1.5 35B-A3B on Mac: coding default and limits2026-10-20T00:00:00Z2026-10-28T00:00:00ZBridge1

Choose Ornith 1.5 35B-A3B when you want Bridge1’s default for coding and have memory for the selected profile. Its published Claude Code result is 40 of 40 small-repository tasks; that supports trying it on bounded coding work, not assuming it will finish every job in your own project.

https://bridge1.ai/resources/models/qwen3-6-35b-a3b-mac/Qwen3.6 35B-A3B on Mac: memory and coding setup2026-10-19T00:00:00Z2026-10-28T00:00:00ZBridge1

You can connect Qwen3.6 35B-A3B to OpenCode, Codex or Claude Code through Bridge1 on an Apple Silicon Mac. Choose the MLX mxfp4 profile when you have room for a growing coding chat: the download is 19.3 GB, and the catalog’s 55.2 GB figure at 229,376 tokens is an estimate for the 64 GB M1 Max.

https://bridge1.ai/resources/models/qwen3-8-27b-mac/Qwen3.8 27B on Mac: choose between two profiles2026-10-19T00:00:00Z2026-10-28T00:00:00ZBridge1

Choose Qwen3.8 27B MLX 4-bit when you want a dense model and can budget for its estimated memory at the reference Mac’s 131,072-token window. Choose GGUF UD-Q4_K_S when you need its 262,144-token admitted window and can work with a 41.8 GB memory estimate rather than a measured peak.

https://bridge1.ai/resources/models/ternary-bonsai-2-27b-mac/Bonsai 2 27B on Mac: small download, larger footprint2026-10-19T00:00:00Z2026-10-28T00:00:00ZBridge1

Ternary Bonsai 2 27B is a small download, but its 8.6 GB file size does not mean an 8.6 GB memory requirement. Choose it when you want compressed Qwen3.8-derived weights and can budget for the catalog’s estimated 53.0 GB at 81,920 tokens on the reference 64 GB M1 Max.