← All articles

Sep 24, 2026 · 10 min read · AI Model Comparison

Jev vs Laya: $40M API vs 421M Open Weights — The Decision-Model Showdown

JEVLAYALAYA MLX1GB RAM AIOPEN SOURCETRENDING 2026
JEVTypeSafe · hosted APIstate + questions→ decision 0.92closed · pay per callLAYAConvai · open weightsstate + questions→ decision 0.88Apache 2.0 · self-hostVS

September 2026 gave us two System One decision models, not one. Jev (TypeSafe AI, San Francisco) is the closed hosted API that crashed servers with demand. Laya (Convai Innovations, India) is the Apache 2.0 open-weights answer — same Choice/Score/Noul primitives, downloadable checkpoints, pip install laya. Same questions, opposite philosophies. Full breakdown below.

IN THIS GUIDE
  1. What each model is
  2. How each works: same primitives, different engines
  3. Head-to-head table
  4. Benchmark chart: 0.766 vs 0.727
  5. Cost chart: API vs GPU
  6. Where to use which
  7. Setup steps for both
  8. Verdict

What each model is

Jev — the hosted decision API

Launched 15 Sep 2026 by TypeSafe AI (Diogo Almeida, ex-OpenAI; $40M seed). Transformer-based, trained with RLCD on synthetic data. You call the API with state + typed questions; it returns decisions with calibrated confidence in 70–500ms. Closed weights, pay-per-call ($0.042/M input, output free), early access clearing fast.

Laya — the open decision weights

Released ~18 Sep 2026 by Convai Innovations (India, CEO Nandakishor M). ModernBERT-large encoder (395M) fully fine-tuned + a decision head trained from scratch (2 transformer layers, option-marker scorer, act/escalate head) — 421M total, Apache 2.0 on Hugging Face. Three checkpoints: laya (English, 512 ctx), laya-multilingual (322M, 1024 ctx, 100+ languages), laya-typed-decisions (fine-tuned specialist, 1024 ctx). Python package with a Router that picks checkpoints per request. ~33ms per question on a T4 GPU.

How each works: same primitives, different engines

Both speak the same language — Choice (pick one + probabilities), Score (ordinal level + distribution), Noul (yes/no probability) — evaluated in one parallel forward pass, no text generated, nothing to parse. The engines differ:

JevLaya
ArchitectureUndisclosed transformer (suspected open-weight base)ModernBERT-large + custom decision head, fully disclosed
TrainingRLCD on synthetic dataRLCD + human-annotated data, TD(λ) for dialogues
Options mechanismUndisclosedEach option scored at its own [MASK] token, softmaxed
Context per questionLarge (API-side)512 tokens (1024 on two checkpoints)
Where it runsTypeSafe serversYour GPU/CPU, offline capable

Head-to-head table

JevLaya
AccessHosted API, waitlist clearingHugging Face + pip install laya
LicenseProprietaryApache 2.0
MultilingualUndisclosed100+ languages checkpoint + Router
Zero-shot qualityStrong generalist (0.727)Near chance — needs fine-tuning
Fine-tuned qualityN/A (can't fine-tune)0.766, beats Jev + teacher ceiling
CalibrationECE 0.144 as shippedOver-confident until you fit temperatures (→0.081)
Data privacyState goes to TypeSafeNever leaves your box
LimitsRate limits, API dependence512–1024 ctx, ~20 options max per question

Benchmark chart: typed-decisions (2,000 decisions)

Accuracy — higher is better (independent benchmark)Laya fine-tuned0.766Teacher ceiling0.735Jev 1.130.727Laya base (zero-shot)~0.36

Read it honestly: fine-tuned Laya beats Jev — but only after training on the benchmark's own split. Base Laya zero-shot is near chance. Jev's strength is working out of the box; Laya's is becoming better than Jev once specialised. Latency: Laya ~33–158ms self-hosted vs Jev ~710ms via API.

Cost chart

1M decisions/monthJev APILaya self-hosted
Unit cost~$0.042/M input, output free$0 marginal
Monthly total~$40–400$100–500 GPU (T4-class, shared)
Fixed costsNoneGPU + ops + fine-tune time
Winner when…Bursty load, no GPU teamSteady load, privacy needs, owned GPUs

Yes, Laya runs on 1GB RAM — all the ways to run it local

This is Laya's killer feature Jev can never match: the 421M weights fit almost anywhere. Community ports cover every runtime:

RuntimeRAM neededSpeedBest for
MLX (Mac) — laya-mlx<1GB7–14ms, 60–75 decisions/sec (Snake demo)MacBooks, background agents
ONNX CPU — int8 quant~0.5GB~64–169ms on 4 CPUsAny laptop, VPS, tight boxes
ONNX / CUDA — fp32~1.7GB~33–38ms on T4Workstations, servers
C++ (ggml)<1GBCPU-nativeEmbedded, no-Python boxes
Node.js (ONNX)~1GBCPUJavaScript stacks
Core ML / Elixir<1GBNeural EngineiOS apps, BEAM systems

No PyTorch, no GPU, no API key, no data leaving the box. Try that with a hosted API.

Where to use which

Pick Jev when…

Pick Laya when…

Calibration warning (both)

Jev ships well-calibrated (ECE 0.144). Laya ships over-confident — fit one temperature per question type on held-out data first (drops ECE 0.466 → 0.081). Never trust raw probabilities for auto-escalation thresholds until measured.

Setup steps for both

Jev in 4 steps

  1. Join early access at TypeSafe AI (~$5 starter credits).
  2. Define ONE decision with 3–5 fixed categories.
  3. Call the API (or LangChain TypeSafeClassifier); log confidences for a week.
  4. Auto-act above 0.95, escalate the rest; measure 30 days.

Laya in 4 steps

pip install laya
# Python:
import laya
agent = laya.load("convaiinnovations/laya")
result = agent.predict(state, questions)  # choice/score/noul + confidence
  1. Evaluate on YOUR data (base is near-chance zero-shot — fine-tune with the notebook on free Kaggle T4s).
  2. Fit calibration temperatures on held-out data.
  3. Deploy behind your API; keep a Jev key as fallback for overflow.

Verdict — the core difference in one paragraph: Jev is a finished product: strong zero-shot (0.727), calibrated out of the box, 64k context, no option cap — but closed, metered, and your data leaves the building. Laya is raw material: near-chance zero-shot (0.362), over-confident until you fit temperatures — but open, free, 1GB-RAM friendly, and fine-tunable past Jev itself (0.766). Jev for deciding this week; Laya for owning decisions forever.

FAQs

Jev vs Laya — which is better?

Fine-tuned Laya wins on accuracy (0.766 vs 0.727); Jev wins on zero-setup convenience. Pick by your constraints, not the leaderboard.

Is Laya free?

Yes — Apache 2.0 weights + pip package. You pay GPU time, not per-call fees.

Can I run Laya locally?

Yes — 421M params runs on a single GPU or CPU. Multilingual checkpoint is smaller (322M).

Does Laya work in Hindi/Punjabi?

Use laya-multilingual (100+ languages) — but measure on your language before trusting it; the English checkpoint collapses on other scripts.

Keep reading:

  1. Jev: The Model That Refuses to Write a Single Word
  2. 5 AI Automations Every Small Business Needs

Want decision models for YOUR industry? Tell us your workflow — we'll map where Jev or Laya fits, including Hindi/Punjabi workloads. Free scoping.

Ask for My Industry → or use the contact form →