Find published research by company, product, platform or technology.
Showing 1–20 of 308 dossiers
Hy4 preview is a very large sparse model with public full and FP8 weights, native speculative decoding and a 1M-token context path. Its open release makes Tencent’s claims testable, while the 1.56TB full checkpoint keeps self-hosting firmly in server-scale territory.
GLM-5.3-Flash combines open weights, multimodal coding/agent capability and an 18B-active sparse architecture with a large anonymous pre-launch trial. Z.ai has already issued a chat-template correction for early downloads, showing that day-one self-hosted deployments need artifact-level validation as well as model benchmarking.
The staged release is complete: GLM-5.3’s public weights and serving artifacts are now available. That makes Z.ai’s coding and cyber-capability claims independently testable while turning the earlier safety delay into a concrete self-hosting and audit decision.
Railway Cloud Agents are managed, persistent development machines rather than a new model or harness. They reuse developers’ existing agent credentials, sleep when disconnected by default, retain disk state, and live inside Railway project environments—blurring the boundary between remote coding workspace and deployment platform.
Beam makes a large agent-focused model commercially reusable without a bespoke licence, but the open weights are not a small local download.
The post-release evidence sharpens the original story. Qwen3.8-27B can retain useful agentic-coding performance at practical 4-bit sizes, but local model quality is not a property of the checkpoint alone: quantization, reasoning effort, context handling and the agent harness can materially change the result.
Kimi K3.1 Pro ships via API at $2 input/$10 output per million tokens, with a 2–3x faster Highspeed tier. Independent Artificial Analysis testing scores 55 on its Intelligence Index and 67% on Terminal-Bench 4.0. Open weights remain promised, not released.
Europe's new large-model contender is available to test through Mistral Studio, but its open-weight promise, pricing and benchmark claims need careful qualification.
MiMo-V2.6 is more useful than another benchmark launch because builders get both capable multimodal weights and a rare view into the reinforcement-learning machinery that produced them: code, environments, run costs and even failure notes from the training cluster.
The useful shift is not another CLI convenience. A coding agent can now create a Shopify dev environment, populate it with existing API and bulk-operation tooling, test against it and tear it down without a person opening the Dev Dashboard.
The useful part of Smaug Agentic is not another frontier-style benchmark claim. Abacus.AI is publishing a drop-in Kimi K3 derivative that targets a specific production failure mode in coding agents: long runs that burn the reasoning budget without converging. The weights and model card are public, but the training data is not disclosed and the benchmark gains remain vendor-produced.
The important shift is that agent orchestration itself becomes a managed API surface: context compaction, tool discovery, programmatic tool calls and subagent coordination can now come from OpenAI’s maintained Codex harness rather than an application team rebuilding those layers.
Android Bench 2.0 moves coding-agent evaluation away from small repository fixes toward dependency upgrades, app builds, migrations and other jobs that can take a human engineer days. The results expose a much larger reliability gap than short-task benchmarks—and show that the agent harness can materially change cost and outcome.
The important failure is not another prompt injection. Plugin4Shell breaks the mechanism intended to guarantee that an AI-agent plugin is still the exact code a marketplace reviewed.
Funes treats agent memory as user-owned data rather than a hosted account feature: retrieval and reranking run locally, provenance stays attached to recalled passages, and cross-machine sharing is optional. The main risk is that publishing session-derived memory can still expose secrets if redaction misses them.
The useful part is not the 800,000-line headline. GitHub has published unusually detailed receipts for a production-scale agent-assisted migration: roughly $120,000 of token spend, 14.5 weeks of incremental releases, dozens of regressions, extensive compatibility tests and a workload-specific jump from 7.55 to 120 session lifecycles per second.
OpenAI’s internal data turns “agents make researchers faster” into a measurable operating model: heavy concurrent agent use, record experiment throughput and rising task complexity, alongside high token spend and persistent human intervention on longer work.
Bounded decision models are turning into a real model category. Cloudflare's entry is open-weight, multimodal and Jev-API compatible, while its fastest variant is aimed at latency-sensitive agent routing.
The release is more interesting than another Qwen3.8 size point because Qwen is deliberately exposing the next architectural generation early. QSA sparse attention, gated residual streams and offloadable n-gram embeddings are now testable before the full Qwen4 family arrives.
Small SaaS teams can now move supported Paddle setup and billing tasks into Claude Code, Codex and similar agent workflows, but production account writes still require careful access control; native metered billing is not generally available.