Tracer (YC S26)

Today we’re launching Echo: Fable-level quality at ~1/3 the inference cost using open-weight models. Echo matches or nearly matches Fable on math, science, and the coding benchmarks we currently cover, while trailing on multilingual and broader knowledge. 🧵

Try Echo: echo.tracerml.ai Read the full Launch YC post: ycombinator.com/launches/SBl-t… Running meaningful LLM inference in production? Work with us to optimize Echo for your workload: cal.com/adam-rida-trac…

Every company’s AI traffic has a shape. For teams that work with us, Echo can be configured around their workload, quality bar, and data requirements to optimize inference further. Public Echo does not train on users’ private traces.

Today we’re launching Echo: Fable-level quality at ~1/3 the inference cost using open-weight models. Echo matches or nearly matches Fable on math, science, and the coding benchmarks we currently cover, while trailing on multilingual and broader knowledge. 🧵

Tracer just launched on @ycombinator's Launch YC! Tracer: Fable-level AI at 1/3 the cost using open-weight models. Check them out: ycombinator.com/launches/SBl-t…

Today, we’re introducing Echo: one adaptive model built entirely from a pool of open-weight models. On our first internal task mix, Echo reached Fable-level results at roughly 1/3 of the total inference cost. Try it: echo.tracerml.ai

@atomicagent_io @eth_chainId

First local agent to beat Hermes on benchmarks! ✦ runs Qwen, Gemma, Llama via llama.cpp ✦ stable-prefix caching keeps sessions cheap ✦ TurboQuant cuts the KV-cache 6.4× smaller ✦ 37 tasks solved vs Hermes' 31 on GAIA Level 1 Open source on macOS, Windows & Linux 👇

@eth_chainId 🔥

@TracerML Insane!!

We’re releasing Echo now to learn from real workloads and keep pushing the approach forward. Full evaluation methodology, results, and limitations: echo.tracerml.ai/eval

You can use Echo through the chat or plug its OpenAI-compatible endpoint directly into OpenCode and other compatible tools. New accounts include $10 in free inference credits. Try it on something hard: echo.tracerml.ai

Tracer 0.2.0 is out New: tracer scan reads your LLM classification traces and tells you, on held-out data it never saw, how much traffic a near-free model can already take over, with an interactive 3D map of your embedding space and more... 👇🧵

@FabraixHQ very cool!

Introducing Nyx, the world’s frontier hacker of AI agents. On AgentHarm, Nyx achieved a 78% attack success rate - compared with 67% for GPT-5.6 Sol. Here’s more about how it works 🧵