Ana Martinez

growth @Celeris_ai | making milliseconds famous. former CPG founder, recovering em dash user

[Ana Martinez @anacmartinez29]

If your decision model takes 127 ms, you built your product around waiting.

[Celeris @Celeris_ai]

@ArtificialAnlys benchmarks Celeris-1 Decision as the fastest decision model at 62ms median response time, 2x faster than @typesafeai Jev.

According to the benchmark, Celeris-1 Decision scores 4 points higher than @OpenAI's GPT 6 Luna.

[Ana Martinez @anacmartinez29]

@Celeris_ai @ArtificialAnlys @typesafeai @OpenAI Latency is going to become a much bigger part of the decision model conversation. We've been thinking about it for a while.

[Ana Martinez @anacmartinez29]

@Celeris_ai @jn2clark So proud to work on this. Speed is one of the big unsolved problems in AI

[Ana Martinez @anacmartinez29]

@jn2clark Great episode! Proud to be part of a team working on one of the most important problems in AI

[Ana Martinez @anacmartinez29]

@oanaolt Greens (amazing view) or Camino Alto!!

[Ana Martinez @anacmartinez29]

@VeraYugen @Celeris_ai send the snack list. i will post the answer and the reason

[Ana Martinez @anacmartinez29]

@TarlonKhoubyari Yesss ✨

[Ana Martinez @anacmartinez29]

the most accurate decision model today did not come from the lab you would have guessed. this will keep happening

[Celeris @Celeris_ai]

Celeris-1 Decision is live today.

On jev-bench and typed decisions it beats Jev 1.13.0, OpenAI gpt-6-luna, Perplexity pplx-decider-v1.1-27b, and Inception Mercury Decide, on both accuracy and Brier score. Celeris-1-decision also led the benchmark in response time with a median of 67ms.

$0.04 per million input tokens, output free.

Full results and methodology: https://t.co/k6gH6SoMx9

[Celeris @Celeris_ai]

Celeris-1 Decision is live today.

On jev-bench and typed decisions it beats Jev 1.13.0, OpenAI gpt-6-luna, Perplexity pplx-decider-v1.1-27b, and Inception Mercury Decide, on both accuracy and Brier score. Celeris-1-decision also led the benchmark in response time with a median of 67ms.

$0.04 per million input tokens, output free.

Full results and methodology: https://t.co/k6gH6SoMx9

[Ana Martinez @anacmartinez29]

Accidentally said spreadsheet instead of GTM engine and they kicked me out of SF

[Tina Feng @tinafenginsf]

Accidentally said marketing instead growth and they kicked me out of SF

[Ana Martinez @anacmartinez29]

@caseykcaruso woww insane view

[Ana Martinez @anacmartinez29]

@Celeris_ai turns out this is what 188ms looks like

[Ana Martinez @anacmartinez29]

benchmarks have entered their contact sport era

[Celeris @Celeris_ai]

We turned model latency into a fighting game. Eight models, single elimination, every frame driven by measured response times.

Celeris-1-Magnus beat GPT-5.6 Sol, Terra and Luna, plus three Gemini Flash models, to take the cup.

The only model that pushed it to the final was Celeris-1, at 188ms.

[Ana Martinez @anacmartinez29]

@GodservantBush @venturetwins https://t.co/bSEF0VCfUP Fast as hell, cheap, and pretty damn good

[Ana Martinez @anacmartinez29]

@tom_w_hamer @huggingface Says not safe for work, therefore must be not safe for work

[Tom Hamer @tom_w_hamer]

Celeris-1 Magnus scored higher than GPT-5.6-sol on τ³-bench, while maintaining faster speed than Luna.

[Ana Martinez @anacmartinez29]

@tom_w_hamer This is where speed starts to actually change what you can build

[Ana Martinez @anacmartinez29]

@jn2clark The team really did cook 🔥

[Ana Martinez @anacmartinez29]

@Celeris_ai If you’re building agents, this is worth trying.

[jesse @jn2clark]

Celeris-1 Magnus is out! The team has been cooking. Built for fast and complex tasks. Hitting well over 500 tok/sec for many use cases while being competitive at the frontier of intelligence.