{
  "schema_version": "1.1",
  "id": "atlas-open-models",
  "slug": "open-models",
  "title": "Open Models",
  "url": "https://feed7.dev/atlas/open-models",
  "current_answer": null,
  "implementation_consequence": null,
  "agent_context": null,
  "confidence": "auto_collected",
  "last_verified": null,
  "last_updated": null,
  "evidence": [],
  "conflicting_sources": [],
  "superseded_claims": [],
  "corpus_evidence": [
    {
      "title": "Data Quality Is the Compute Multiplier — Ari Morcos, DatologyAI",
      "url": "https://www.youtube.com/watch?v=_PdK6x7PQNM",
      "source_name": "AI Engineer",
      "published_at": "2026-07-31T23:00:06+00:00",
      "summary": "Training-data curation can improve model quality and inference efficiency without simply adding compute. The practical work is decontamination, deduplication, balancing, task matching, and selective synthesis."
    },
    {
      "title": "Agents at Scale: Inside MiniMax's Model and the Infrastructure Behind It — Olive Song",
      "url": "https://www.youtube.com/watch?v=AVMr9PMINyo",
      "source_name": "AI Engineer",
      "published_at": "2026-07-31T03:00:14+00:00",
      "summary": "MiniMax and Together describe why agent workloads change inference priorities: large codebase context, repeated tool turns, multimodality, and long-horizon RL all demand purpose-built environments and kernels."
    },
    {
      "title": "$β$-OPSD: Deriving with Policy Optimization, Training with Self-Distillation",
      "url": "https://arxiv.org/abs/2607.28582v1",
      "source_name": "arXiv",
      "published_at": "2026-07-30T17:41:16+00:00",
      "summary": "β-OPSD exposes self-distillation’s fixed regularization as a tunable parameter, then approximates policy optimization through logit mixing. It targets more stable reasoning training without direct RL."
    },
    {
      "title": "Serving 2 Million Models Without Melting: Scaling the Hugging Face Hub — Arek Borucki, Hugging Face",
      "url": "https://www.youtube.com/watch?v=lyL5QhgIOxc",
      "source_name": "AI Engineer",
      "published_at": "2026-07-28T13:41:11+00:00",
      "summary": "Hugging Face’s Hub shows how to scale model search: precompute tokens, separate metadata from artifacts, isolate heavy reads, and autoscale from workload signals."
    },
    {
      "title": "DataOrchestra: Learning to Orchestrate Per-Example Curation of Pretraining Data",
      "url": "https://arxiv.org/abs/2607.24717v1",
      "source_name": "arXiv",
      "published_at": "2026-07-27T17:54:12+00:00",
      "summary": "DataOrchestra chooses a processing pipeline per pre-training example, improving average benchmark results while avoiding compute on chunks that need no transformation."
    },
    {
      "title": "The Messy Reality of Scale: Synthetic Data and Pre-Training — Marah Abdin & Robert McHardy, poolside",
      "url": "https://www.youtube.com/watch?v=KhYifX22yhE",
      "source_name": "AI Engineer",
      "published_at": "2026-07-26T01:00:06+00:00",
      "summary": "poolside treats synthetic data and training observability as one scaling problem: diversify scarce code data, decompose hard generation tasks, and verify numerical invariants before failures compound."
    },
    {
      "title": "Why Large? Tiny LMs & Agents on Edge/Robotics — Cormac Brick, Google",
      "url": "https://www.youtube.com/watch?v=hacEQHHhu2Q",
      "source_name": "AI Engineer",
      "published_at": "2026-07-25T17:00:06+00:00",
      "summary": "Edge-agent design is often memory-bound: use small prompted models when devices can spare gigabytes, or fine-tune tiny models for narrow functions when reach and latency matter more."
    },
    {
      "title": "Training Frontier Models to Out-Think Hackers — Uri Rolls, Arithmetic & Thom Wolf, Hugging Face",
      "url": "https://www.youtube.com/watch?v=O-CBZ3JtRvo",
      "source_name": "AI Engineer",
      "published_at": "2026-07-24T05:19:15+00:00",
      "summary": "This security eval tests whether agents can discover and exploit logic flaws across live chained services, using hidden zero-days and deterministic grading instead of source-code pattern matching."
    },
    {
      "title": "3D-Aware VLMs with Implicit and Explicit Geometries",
      "url": "https://arxiv.org/abs/2607.21595v1",
      "source_name": "arXiv",
      "published_at": "2026-07-23T17:59:59+00:00",
      "summary": "VLM-IE3D adds implicit and reconstructed geometry tokens to an RGB-video VLM, offering an open approach for agents that must reason about spatial scenes without dedicated 3D input."
    },
    {
      "title": "Notion's Token Town — Sarah Sachs, Notion",
      "url": "https://www.youtube.com/watch?v=-I5W5QVAT8E",
      "source_name": "AI Engineer",
      "published_at": "2026-07-23T17:30:06+00:00",
      "summary": "Agent economics can regress even when token prices look stable. Route by task, preserve model optionality, and move deterministic work out of LLM calls before scaling usage."
    },
    {
      "title": "Ling 3.0 Flash is now available on AI Gateway",
      "url": "https://vercel.com/changelog/ling-3-0-flash-is-now-available-on-ai-gateway",
      "source_name": "Vercel",
      "published_at": "2026-07-23T00:00:00+00:00",
      "summary": "Ling 3.0 Flash joins AI Gateway with a 256K context window, thinking and non-thinking modes, and free access through August 3 for agent workload testing."
    },
    {
      "title": "Laguna S 2.1 is now available on AI Gateway",
      "url": "https://vercel.com/changelog/laguna-s-2-1-is-now-available-on-ai-gateway",
      "source_name": "Vercel",
      "published_at": "2026-07-21T00:00:00+00:00",
      "summary": "Poolside’s open-weight Laguna S 2.1 is now callable through Vercel AI Gateway in free 256K and paid 1M-context variants, with thinking and no-thinking modes for coding agents."
    },
    {
      "title": "Hugging Face and Cerebras bring Gemma 4 to real-time voice AI",
      "url": "https://huggingface.co/blog/cerebras-gemma4-voice-ai",
      "source_name": null,
      "published_at": null,
      "summary": "Hugging Face demos real-time speech-to-speech from open parts: Nvidia Parakeet ASR, Gemma 4 31B on Cerebras inference, Alibaba's Qwen3TTS — pipeline code is open and already runs on 9,000+ Reachy Mini robots."
    },
    {
      "title": "Program-as-Weights: A Programming Paradigm for Fuzzy Functions",
      "url": "https://arxiv.org/abs/2607.02512v1",
      "source_name": null,
      "published_at": null,
      "summary": "Program-as-Weights compiles natural-language fuzzy functions (JSON repair, log filtering) into adapters for a frozen 0.6B interpreter — matching Qwen3-32B prompting at ~1/50th the memory, 30 tok/s on an M3."
    },
    {
      "title": "Zackriya-Solutions/meetily",
      "url": "https://github.com/Zackriya-Solutions/meetily",
      "source_name": null,
      "published_at": null,
      "summary": "Local-first meeting assistant: Whisper/Parakeet transcription with a 4x speed claim and Ollama summaries, all on-device via a Rust/Tauri app — a working template for shipping private, local AI features."
    },
    {
      "title": "CompactionRL: Reinforcement Learning with Context Compaction for Long-Horizon Agents",
      "url": "https://arxiv.org/abs/2607.05378v1",
      "source_name": null,
      "published_at": null,
      "summary": "CompactionRL uses RL to teach agents to compact their own context mid-task, lifting GLM-4.5-Air 7 points to 66.8% on SWE-bench Verified; the recipe is now in GLM-5.2's training pipeline."
    },
    {
      "title": "The latest AI news we announced in June 2026",
      "url": "https://blog.google/innovation-and-ai/technology/ai/google-ai-updates-june-2026/",
      "source_name": null,
      "published_at": null,
      "summary": "Google's June roundup: Gemma 4 12B runs locally in 16GB of memory, Gemini 3.5 Flash adds computer use for desktop, mobile, and browser agents, and Nano Banana 2 Lite ships as a cheaper image model."
    },
    {
      "title": "Lordog/dive-into-llms",
      "url": "https://github.com/Lordog/dive-into-llms",
      "source_name": "GitHub",
      "published_at": null,
      "summary": "A free, code-oriented Chinese curriculum spans model tuning, deployment, agents, alignment, security, and multimodal systems. It is useful as a broad learning map, but remains a work in progress."
    },
    {
      "title": "Open-weight models surge to 29% of volume, price per token flattens",
      "url": "https://vercel.com/blog/ai-gateway-production-index-july-2026",
      "source_name": null,
      "published_at": null,
      "summary": "Vercel’s June gateway data shows cheap volume moving to open-weight models while costly agent workloads stay on frontier models, strengthening the case for risk-based routing."
    },
    {
      "title": "Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel",
      "url": "https://huggingface.co/blog/nvidia/accelerating-fine-tuning-nvidia-nemo-automodel",
      "source_name": null,
      "published_at": null,
      "summary": "NVIDIA's NeMo AutoModel makes MoE fine-tuning ~3.4-3.7x faster on Transformers v5 while keeping the AutoModelForCausalLM API — relevant if you tune open models like Qwen3-30B-A3B for your own agents."
    },
    {
      "title": "Run a vLLM Server on HF Jobs in One Command",
      "url": "https://huggingface.co/blog/vllm-jobs",
      "source_name": null,
      "published_at": null,
      "summary": "HF Jobs now stands up an OpenAI-compatible vLLM endpoint in one command, billed per second (A10G at $1.50/hr). Useful for throwaway endpoints: one-off evals, batch runs, agent experiments against open models."
    }
  ]
}