We're merging code edits at over 10,500 tokens/sec now. Thats 4x faster than the fastest model on cerebras and 175x faster than Claude Sonnet 4
Morph
Software Development
San Francisco, California 562 followers
Fast, Reliable AI Edits - 4500 tok/sec
About us
Models for building the best coding agents. The best search, edits, and more
- Website
-
https://morphllm.com
External link for Morph
- Industry
- Software Development
- Company size
- 2-10 employees
- Headquarters
- San Francisco, California
- Type
- Privately Held
Locations
-
Primary
Get directions
San Francisco, California 94109, US
-
Get directions
1355 Leavenworth St
San Francisco, California 94109, US
Employees at Morph
Updates
-
Morph reposted this
Our code writing success rate jumped from 82% to 96%. All from a single update. This is how: We integrated Morph (YC S23) into our backend, a model trained to update existing code. Their Fast Apply model merges AI-generated code at 4500+ tokens/sec. With 98% accuracy. This is achieved through its semantic understanding of code, aligning it with the problem at hand (not just pattern matching like general models). This means less failed builds and more clean code. Picking the right model for the right tasks can have significant impact. For anyone building their own tools, go experience clean code building in Databutton. For any team building AI-assisted development, go check out Morph 👍 BTW: Not paid or asked to post this, just a happy customer :)
-
-
Morph reposted this
Our code writing success rate jumped from 82% to 96%. All from a single update. This is how: We integrated Morph (YC S23) into our backend, a model trained to update existing code. Their Fast Apply model merges AI-generated code at 4500+ tokens/sec. With 98% accuracy. This is achieved through its semantic understanding of code, aligning it with the problem at hand (not just pattern matching like general models). This means less failed builds and more clean code. Picking the right model for the right tasks can have significant impact. For anyone building their own tools, go experience clean code building in Databutton. For any team building AI-assisted development, go check out Morph 👍 BTW: Not paid or asked to post this, just a happy customer :)
-
-
Morph reposted this
Morph (YC S23) is what AI coding agents need to be world-class at retrieving and editing code. Fast Apply, embeddings, and reranking models - built for code. Agents fail today because they retrieve the wrong context, or they write changes that break the code (diff-patch chaos). Morph gives companies a drop-in API to get top-tier coding performance: 1) Syntax-aware Embeddings → the exact context your LLM needs, every time. 2) Reranker → strips out everything else. 3) Fast-Apply model → applies the AI suggested edit into your code at 1600+ tok/sec. 8 × fewer errors than patch diffs Drop-in API or full on-prem. The Morph team is made up of ex-Tesla engineers who've trained models at the largest scale. They're building Morph so teams building coding agents can focus on building unique features their customers love. If you're building Cursor-for-X, an autonomous PR bot, or in-house code assistants, reach out to founders@morphllm.com and try their APIs for free at https://morphllm.com. Congrats on the launch, Tejas Bhakta and team! https://lnkd.in/gCnCCMPX