Applied Compute developed a 35B open-weight model trained to search a precomputed index answering repo search questions at 100x lower cost than a frontier model.
Public source
Publisher name
Public post
A 35B open-weight model trained to search a precomputed index answers repo search questions at 100x lower cost than a frontier model. We partnered with turbopuffer to tr…
Company
Applied Compute
- Industry
- Technology, Information and Internet
- Location
- San Francisco, US
- Company size
- 11–50 employees
About Applied Compute
Applied Compute builds Specific Intelligence for enterprises, unlocking the knowledge inside a company to train custom models and deploy an in-house agent workforce.
See moreLatest activity
Latest activity from Applied Compute
4 signals
Products & Services
Applied Compute Agent Cloud now turns post-training into a repeatable, engineered process using harnessing multi-trillion-parameter models with latest research techniques.
Partnerships
Applied Compute partnered with turbopuffer to train Qwen3.6-35B-A3B to find code across ~9,000 repositories.
Research & Knowledge
Applied Compute CEO Yash Patil sat down with Andy Fang to discuss why cheaper software doesn’t erase years of operating advantage and the moat of DoorDash.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
LlamaIndex
LlamaIndex evaluated 14 frontier systems across 370 enterprise documents to plot accuracy against cost per page.
Research & Knowledge
Magic
Magic developed a new pretraining recipe for Frontier that matches DeepSeek V4 Pro's pretrain using 50x less compute, roughly half the FLOPs used for GPT3, or ~$0.5M on GB200.
Research & Knowledge
Concentrate AI
Concentrate AI reports that 1 billion tokens on GLM-5.3-Flash costs ~$35, which is 26x cheaper than GPT-5 High.
Research & Knowledge
GitHub
GitHub reached frontier-level coding quality in GitHub’s offline benchmarks while reducing estimated workflow cost by up to 67% versus Claude Opus 5 on one benchmark.
Research & Knowledge
The New Stack