Jev Mod for Claude Code: Saving Tokens and Accelerating Skill Handling
Developers introduced the Jev mod for Claude Code, which optimizes AI context window usage by loading skills on demand and speeding up request processing by several times.
A new mod for Claude Code has appeared, significantly saving tokens. The key feature of Jev is that instead of loading all available skills into the model's context window immediately, they remain outside of it until they are actually needed.
Jev analyzes the available skills and selects the appropriate one for each request, dynamically adding it to the context. This allows avoiding unnecessary load on the AI assistant's context window.
Test Results
The team tested Jev Skill Suggestion against various Claude models. The experiment included 26 skills, 10 prompts, and 3 runs per prompt. The data obtained:
- Jev: 90% accuracy, median response time — 330 ms
- Claude Haiku: 80% accuracy, 1.24 s
- Claude Sonnet / Opus / Fable: 100% accuracy, 2.2–4.5 s
Conclusions
Jev slightly sacrifices absolute accuracy (90% vs 100% in flagship models) but makes decisions 4–14 times faster. The main advantage is that the list of skills does not occupy valuable space in the context window at all, making work with a large number of tools more efficient.
Why it matters
AnalysisContext window optimization is critical for complex agentic systems where the number of available tools is constantly growing. Jev demonstrates how to balance speed and accuracy without overloading the AI with unnecessary data.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Jev and System One AI: A Complete Practical Guide to the Decision Model from TypeSafe
A deep engineering breakdown of the flagship Jev model: System One architecture versus LLMs, Choice/Score/Noul primitives, request parallelism, confidence mathematics, video use cases, code, and an overview of local open-weight alternatives (Laya, GLiNER2.5, CLM-8B).
Jev Decision Model
Jev is a commercial non-autoregressive System 1 model from TypeSafe AI for fast typed decisions in agentic loops: it returns a validated decision object in 70–500 ms at ~$0.042/1M input tokens, eliminating JSON syntax hallucinations and high-latency routing in ReAct loop/Tool Calling stacks.
OpenAI Frees Several Gigabytes of RAM in ChatGPT App
A ChatGPT app update reduces memory use in long local sessions: after 32,000 responses, RAM consumption stays around 1.7 GB instead of up to 6 GB.
Migrating from Claude Code to Codex: How to Move CLAUDE.md, MCP, Skills, Slash Commands, and Settings
Complete engineering guide on migrating from Claude Code to OpenAI Codex CLI: porting CLAUDE.md to AGENTS.md, configuring MCP, sandboxing, hooks, and retaining Claude models.