APPLICATION

Laya: a 421M non-autoregressive System 1 decision model at 33 ms per forward pass

conval-laya

Convai Innovations' open-source non-autoregressive System 1 decision model: 421M parameters, ModernBERT backbone with [MASK] option extraction, one 33 ms forward pass per calibrated decision, Apache-2.0; trained with RLCD strictly proper scoring rules for honest probabilities; the model card ships a real benchmark against Jev plus an honest limitations list. Positioned for edge and high-concurrency small decisions (game NPCs, dialogue policy, request routing), not long reasoning or open generation.

CONFIDENCE: Vendor Claim; Official model card or keynote only, no independent re-test
KEY METRIC: Single forward-pass latency (421M, vendor); Vendor Claim · 2026-09
MATURITY: Research; research → demo → product → production

Editor's take
Reading: 421M parameters at 33 ms puts System 1 decisions inside real-time loops, Apache-2.0 weights make reproduction nearly free, and the card's head-to-head against Jev plus its limitations list are high-water marks for open releases. But non-autoregressive option extraction is capped by option-set quality, and RLCD calibration out of distribution still lacks third-party reproduction; using it as a decision co-processor rather than the main model is the correct posture. Confidence C (vendor claim).

System 1 ModelNon-autoregressiveRLCDConvai Innovations

Convai InnovationsOfficial site github.com/NandhaKishorM/laya1 min read672

Laya is Convai Innovations' open-source non-autoregressive System 1 decision model: 421M parameters, a ModernBERT backbone with [MASK] option extraction, one 33 ms forward pass per calibrated decision, weights and code released under Apache-2.0. It does not generate text; it extracts and scores over a given option set, compressing small decisions into a size that fits inside real-time loops.

Training uses RLCD with strictly proper scoring rules: the objective is not human preference ranking but statistically honest output probabilities, so confidences can be consumed as decision risk. That is the deepest split from RLHF-style decision heads and the clearest section of the model card.

The card ships a real benchmark comparison against Jev plus an honest limitations list (option-set coverage, out-of-distribution calibration gaps, multilingual tier differences) - one of the few open releases we have read that names a competitor directly; siblings laya-multilingual and laya-typed-decisions cover variant tiers.

Boundary: edge and high-concurrency small decisions (game NPCs, dialogue policy, request routing), not long reasoning or open generation; option-set quality caps the ceiling, and open-domain questions need a sequential model to write the candidate set first.

More in LLM

SOTA hub

LLMDevelopmentTopC

GLM-5.3: the open-weights coding flagship bought with post-training alone

Z.ai post-trained the very same base as GLM-5.2 and nothing else, producing the 744B-A40B GLM-5.3: 50 percent over 5.2 on the in-house Z.ai Code Bench, open-weights first on Terminal Bench 3.0 at 28.3 and Agents' Last Exam at 28.5, while AutomationBench 48.2 and GDPVal-AA v2 1769 beat every closed-source comparison on the official chart; CyberGym vulnerability discovery is SOTA and the exploitation-chain gains emerged without dedicated training. The companion 320B-A18B Flash starts from a new base and is the first GLM to mix sparse with linear attention, attacking long-context serving cost. Weights ship in FP8 and BF16, deployable on SGLang, vLLM, KTransformers and the Ascend stack.

28.3Terminal Bench 3.0 (top open-weight, vendor chart)Vendor Claim · 2026-08

ProductZ.ai (Zhipu AI)Site Repo

ReasoningDevelopmentTopA

Claude Fable 5.1: the long-horizon agentic tier that tops the ladder on a different ruler

The fixed version Anthropic positions for demanding reasoning and long-horizon agentic work (claude-fable-5-1, retirement no earlier than 2027-09-01): default thinking effort high at $10/$50 per million tokens, 2.5x sibling Opus 5.5, and default-high is itself a default cost behaviour. Its standing only holds once you switch rulers: #1 on net_improvement in the Arena Agent board (13.71, confirmed_success 19.83, praise 31.83) ahead of GPT-6 Astra (11.54/17.70/32.79), while Opus 5.5 misses the top eight; #2 on Terminal-Bench 4.0 at 57.88% (n=330), 0.3 points behind Astra at 58.18% but with pass@5 of 0.7879 against 0.7121, lower single-shot and more robust over five attempts; #4 on the intelligence index at 53.35 (max), below Opus 5.5. The picture is consistent: not first on composite intelligence, first on pushing a real task forward. A separate praise column means the board carries a human or judge component and is not an objective benchmark. Choose by whether the workload is hard from reasoning or from long-horizon consistency. Closed, API only. Graded A (confirmed); no long-task comparison on a standard harness.

13.71Arena Agent net improvementConfirmed · 2026-09

ProductAnthropicSite

ReasoningDevelopmentTopA

GPT-6 Astra: first on terminal engineering tasks, third tier on composite intelligence

The OpenAI flagship tier, officially positioned for hardest end-to-end work (gpt-6-astra): 1.05M-token context, 128K-token single output, knowledge cutoff 2026-04-30, with Functions, Web search, File search and Computer use built in, so search, retrieval and interface control ship with the model instead of an outer agent framework. Reasoning effort runs low to max in five steps while price stays fixed at $10/$50, so the cost lever is token consumption; siblings Sol ($2/$10) and Luna ($0.1/$0.5) make a 100x spread, so load splitting can stay inside one vendor. Readings (2026-09-22): #1 on Terminal-Bench 4.0 at 58.18% (n_trials=330, pass@5 0.7121), the protocol closest to a coding agent's daily life; #2 on Arena Agent with net_improvement 11.54 and confirmed_success 17.70, both below leader Fable 5.1, but the field's highest praise at 32.79; #6 and #7 on the intelligence index (max 52.67). The cutoff is nearly five months older than this page, so version numbers and API changes must go through Web search; computer-use reliability appears on no board; closed, API only. Graded A (confirmed); not independently benchmarked.

58.18%Terminal-Bench accuracyConfirmed · 2026-09

ProductOpenAISite

LLMDevelopmentTopA

MiMo-V2.6: Live RL as an auditable ledger of 30 steps, 750k trajectories, and dollars

Xiaomi's open-source MiMo-V2.6 Pro/Flash: 30 steps of Live RL, roughly 750k trajectories, public cost of $850k/$2.62M, and 7k+ RL environments released in the same batch; tops the open-weights camp on the Artificial Analysis intelligence index at 46 with +17/+14 points out-of-sample on DeepSWE v1.1; four demo lines (Vibe World, CUA, research, content creation) all open. Self-improvement becomes an auditable ledger - steps, trajectories, and dollars on the table for third-party recomputation.

46Artificial Analysis Intelligence Index (top open-weight)Confirmed · 2026-09

ProductXiaomiSite Repo

As an Amazon Associate, we earn from qualifying purchases.