--- title: The full stack goes open: giant weights on serverless, a 30B agent on one GPU, open RISC-V, one x86 description: Meta's 816B open-weights model lands on serverless, a 30B agent model targets a single GPU, Google open-sources RISC-V NPU silicon, and AMD and Intel unify x86 AI work in oneDNN. Openness is running at every layer — all figures vendor-reported. canonical: https://hangingcontext.com/stories/2026-10-21/openness-every-layer/ last-updated: 2026-10-21 url: https://hangingcontext.com/stories/2026-10-21/openness-every-layer/ published: 2026-10-21T12:48:47Z edition: 2026-10-21T12 newsroom: artificial-intelligence beat: models-labs trust: confirmed source_count: 4 cited_manifest_ids: ['1792530891857253442', '1792582719791107060', '1792511537153435644', '1792554626694101468'] --- # The full stack goes open: giant weights on serverless, a 30B agent on one GPU, open RISC-V, one x86 *Meta's 816B open-weights model lands on serverless, a 30B agent model targets a single GPU, Google open-sources RISC-V NPU silicon, and AMD and Intel unify x86 AI work in oneDNN. Openness is running at every layer — all figures vendor-reported.* Confirmed · 4 sources · Artificial Intelligence · Published 2026-10-21 12:48 UTC ## Key facts - Meta's Muse Spark Open — an 816B-parameter mixture-of-experts model with 42B parameters active per token and a 1M-token context window — is available on Fireworks' US serverless platform with zero data retention by default, per the companies. (self-reported) [cites: 1792530891857253442] - Fireworks reports 91.1% on AIME26, 89.7% on GPQA Diamond, 84.3% on SWE-bench Verified, and 82.9% on Terminal-Bench 2.1, plus a 33% hallucination rate on AA-Omniscience — all vendor-reported. (self-reported) [cites: 1792530891857253442] - Muse Glimmer 30B is an Apache 2.0-licensed open-source model for on-device agentic tasks via 4-bit quantization, claimed as state-of-the-art for its size class, described in a preprint this week. (self-reported) [cites: 1792582719791107060] - Google Research completed a 32-bit open-source Coral NPU RISC-V architecture targeting 1 TOPS at 10 mW average power, with a v1.0 open-source IREE/MLIR compiler deploying models from JAX, PyTorch, and TensorFlow. (self-reported) [cites: 1792511537153435644] - AMD and Intel aligned on oneDNN as the default open-source library for x86 AI acceleration within PyTorch, and are collaborating on AI Compute Extensions, a next-generation x86 ISA for matrix multiplication. (self-reported) [cites: 1792554626694101468] ## 816B weights on serverless Meta's Muse Spark Open is now available on Fireworks' US serverless platform. Fireworks describes an 816B-parameter mixture-of-experts model with 42B parameters active per token and a 1M-token context window, built for long-horizon coding and agentic workflows — and says it runs with zero data retention by default. Fireworks pitches it for regulated professional tasks on the strength of the legal and medical benchmark results below. [cites: 1792530891857253442] The benchmark sheet Fireworks published is aggressive: 91.1% on AIME26, 89.7% on GPQA Diamond, and 84.3% on SWE-bench Verified, per the company's post. In Meta's evaluations it scores 82.9% on Terminal-Bench 2.1 — though both sets of figures are vendor-reported. [cites: 1792530891857253442] On professional tasks, Fireworks says the model ranks first of 75 models on Vals AI's Harvey Egal legal-agent benchmark at 25.4%, scores 53.3% on Rogo's Big Finance Bench — the top open-weights result — and hits 90.1% on Vals MedScribe. Fireworks says the model outperformed Claude Opus 5.5 on Rogo's Big Finance Bench. Its hallucination rate on AA-Omniscience is 33%, the company says. [cites: 1792530891857253442] > **From Synorb** — Run the cited Manifests yourself. Every claim in this paper points to a Manifest id. Get a Synorb Key and read the Manifests behind it. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; 3 of them are starter Beacons. [Get a key](https://synorb.com/connect) For regulated industries, the combination Fireworks is selling is open weights plus zero data retention by default — the compliance story that goes with the benchmark sheet. [cites: 1792530891857253442] ## A 30B agent that fits on one GPU Muse Glimmer 30B, described in a paper this week, is a 30-billion-parameter open-source causal language model engineered for agentic tasks on consumer hardware. With a perception encoder and 4-bit quantization, it performs multi-step reasoning and tool use locally on a single GPU — no cloud required — and is released under the Apache 2.0 license. [cites: 1792582719791107060] The authors claim state-of-the-art performance for its size class on agentic benchmarks, including tool use, multi-step planning, and task completion ahead of Gemma4-31B and Qwen3.6-27B. Its perception encoder supports failure recovery locally, keeping the full agent loop on-device. Preprint claims, but the direction is clear: the agent stack is compressing onto hardware people already own. [cites: 1792582719791107060] ## Open silicon Google Research says it has completed a 32-bit open-source Coral NPU RISC-V architecture with a Matrix Execution Unit, delivering up to 1 TOPS while targeting 10 mW average power for edge inference. Alongside it came a v1.0 open-source IREE/MLIR compiler that takes models from JAX, PyTorch, and TensorFlow straight to Coral NPU hardware without custom assembly. [cites: 1792511537153435644] Developers can prototype on the Synaptics Astra SR200 MCU series, and Coralboard devkits are globally available, the company says. The pitch: portable, energy-viable intelligence from model to silicon, with no proprietary toolchain in the middle. [cites: 1792511537153435644] ## One x86 AMD and Intel are combining their x86 CPU optimization work for PyTorch in oneDNN, the open-source deep learning library governed by the Unified Acceleration (UXL) Foundation — following a recommendation from the x86 Ecosystem Advisory Group. AMD is integrating its ZenDNN optimizations into oneDNN and has joined the UXL Foundation as a steering member, the companies said. [cites: 1792554626694101468] They are also collaborating on AI Compute Extensions, a next-generation x86 instruction set architecture designed to accelerate matrix multiplication, with PyTorch integration planned through oneDNN updates. The stated aim is to reduce fragmentation in the x86 AI ecosystem so PyTorch applications are easier to build, maintain, and deploy consistently across x86 systems. [cites: 1792554626694101468] ## What is still unresolved - Benchmark figures for Muse Spark Open come from Fireworks and Meta; no independent replication appears in the evidence. - Muse Glimmer's size-class claims are preprint claims, not independently verified. - AI Compute Extensions is a forecast — silicon and PyTorch integration are not yet delivered. ## Decision card - **Useful for:** Teams choosing open-weights models for agents or regulated workloads; Developers deploying agents on-device or at the edge; Hardware teams evaluating open RISC-V NPUs or x86 AI acceleration - **Not for:** Treating vendor benchmark sheets as independent results; Buying decisions on unreleased silicon or standards - **Compare against:** Closed frontier models' benchmark sheets; Other edge-AI stacks; Existing standalone x86 optimization paths ## Provenance ledger | Manifest id | Title | Source | Claims used | Published | |---|---|---|---|---| | `1792530891857253442` | Muse Spark Open on Fireworks: Meta's open weights wodel, available on US serverless | [fireworks-ai-blog](https://fireworks.ai/blog/meta-muse-spark-open) | - | 10-20 00:00 | | `1792582719791107060` | Muse Glimmer 30B: An On-Device Agentic Model | [huggingface-daily-papers](https://huggingface.co/papers/2610.24628) | - | 10-21 00:00 | | `1792511537153435644` | Unifying Edge AI: How Coral NPU's open standard brings portable, energy-viable intelligence from model to silicon | [google-developers-ai](https://developers.googleblog.com/unifying-edge-ai-how-coral-npus-open-standard-brings-portable-energy-viable-intelligence-from-model-to-silicon) | - | 10-20 00:00 | | `1792554626694101468` | AMD and Intel Collaborate to Simplify AI Development on x86 CPUs Using PyTorch | [intel-blog](https://intel.com/content/www/us/en/newsroom/news/data-center/amd-intel-simplify-x86-ai-development-with-pytorch.html) | - | 10-21 00:00 | ## Run these Manifests yourself Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them. 1. **Get a Synorb Key.** Click Generate Free Agent Key at https://synorb.com/connect, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{"client":"your-agent-name"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; 3 of them are starter Beacons. 2. **Read each cited Manifest by id.** REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer `. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief. 3. **Check the claim.** Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them. Cited Manifest ids: - `1792530891857253442` - `1792582719791107060` - `1792511537153435644` - `1792554626694101468` ```sh for id in 1792530891857253442 1792582719791107060 1792511537153435644 1792554626694101468; do curl -s -H "Authorization: Bearer $SYNORB_KEY" "https://api.synorb.com/manifests/by-id/$id" done ``` > **From Synorb** — Connect your agent over MCP. Add https://mcp.synorb.com/mcp to your agent, then call the tool synorb-get-demo-key if it has no key yet. It needs no email and no form. Keys are free for 72 hours, and new keys are limited per network. [How to connect](https://synorb.com/connect) ## Related stories - [Hallucination gets a taxonomy: five preprints measure where models invent, drift, and go silent](https://hangingcontext.com/stories/2026-10-21/hallucination-taxonomy/) - [The evals are the problem: refusal tests that don't refuse, guardrails that memorize, and one space that moves everything](https://hangingcontext.com/stories/2026-10-21/evals-are-the-problem/) - [The context pipeline: federated data, live mirrors, graph layers, and expertise over MCP](https://hangingcontext.com/stories/2026-10-21/context-pipeline-agents/) - [Benchmarks say AI can't replace programmers yet -- but the power users are shipping anyway](https://hangingcontext.com/stories/2026-10-21/code-agent-benchmark-gap/) ## Links - HTML: https://hangingcontext.com/stories/2026-10-21/openness-every-layer/ - JSON: https://hangingcontext.com/stories/2026-10-21/openness-every-layer/index.json - Markdown: https://hangingcontext.com/stories/2026-10-21/openness-every-layer/index.md Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify. --- Reuse (terms v1): Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product. Credit: "Hanging Context, with provenance by Synorb", with a link to https://hangingcontext.com/stories/2026-10-21/openness-every-layer/. Terms: https://hangingcontext.com/terms/reuse/