walid@portfolio:~/lab/giant-lab-open-releases$
cd../lab
01ideaSep 2026

Seven open releases, two licence traps

What OpenAI, xAI, NVIDIA, Anthropic, Cloudflare, Tencent and Alibaba put on GitHub — and which ones you can actually bill for

Seven things the big labs open-sourced, all of them running on your own hardware rather than someone else’s API. Five are plain Apache 2.0 and you can put them straight into client work. The other two are where the actual question lives: one ships MIT code with separately licensed weights, and one excludes the EU, the UK and South Korea outright.

Open sourceLicensingAgentsVoice3DSymphonyx-algorithmCloudflare OS
i
The licence is the interesting part

Every one of these is an official release from the company that built it, and every one runs on your own machine or your own cloud account — no per-seat API bill sitting between you and a client. That makes the licence the thing that decides whether a given one is a business or a toy, which is exactly the part these round-ups skip. Two of the seven carry real restrictions, and one of them is geographic.

Checked, not copied

The seven, verified against each repository on 20 September 2026

Licences read from the LICENSE file in each repo rather than taken from the write-up.

ReleaseWhat it isLicence
OpenAI — SymphonyWatches a Linear board and spawns Codex agents per task, returning proof of workApache 2.0
xAI — x-algorithmThe ranking code behind the For You feedApache 2.0
NVIDIA — PersonaPlexA 7B full-duplex speech model: listens and talks at onceMIT code, NVIDIA Open Model weights
Anthropic — financial-servicesFinance agents and plugins for Claude CodeApache 2.0
Cloudflare — Cloudflare OSThe internal AI workspace, agents building small appsApache 2.0
Tencent — Hunyuan3D-2One photo to a textured 3D modelTencent Hunyuan 3D 2.0 Community Licence
Alibaba — Qwen3-TTSVoice design and cloning across 10 languagesApache 2.0
!
The two that are not Apache 2.0

Tencent’s licence opens by stating in capitals that it does not apply in the European Union, the United Kingdom or South Korea, and it requires a separate licence from Tencent once a product built on it passes one million monthly active users. If the client is European or British, this one is off the table regardless of what you charge. NVIDIA’s split is gentler but still a split: the PersonaPlex code is MIT, while the weights are under the NVIDIA Open Model licence and gated — you accept the terms on the model page before Hugging Face will hand them over. Everything else here is plain Apache 2.0.

!
The X ranking advice that round-ups keep getting wrong

The usual reading is that negative signals are catastrophic — a report is weighted hundreds of times a like, so one report wipes out hundreds of likes. xAI’s README calls this out directly as a misconception. The weights scale the model’s predicted probability that *you specifically* would take that action, not raw counts of who did. So a high report weight does not mean one report cancels 468 likes; it means the model’s estimate of how likely you are to report something is worth a lot in your own score, and that estimate is driven largely by your own past behaviour. xAI says it added comments to the ranking code precisely so that people and LLMs reading it stop drawing the wrong conclusion — which is worth knowing before you sell strategy built on the wrong mental model.

Read the repos, not the summary

What each one turned out to include

01·OpenAI

Symphony expects a codebase already set up for agents

The README is explicit that it works best in codebases that have adopted harness engineering — it is the step after managing coding agents, not a way to start. Proof of work is richer than the usual summary suggests: CI status, PR review feedback, complexity analysis and walkthrough videos. OpenAI labels it a low-key engineering preview for trusted environments, so every change still gets reviewed before it merges.

02·xAI

Two retrieval paths, then one model ranks both

In-network posts come from Thunder, which keeps recent posts from accounts you follow in memory. Out-of-network comes from Phoenix retrieval and SimClusters. Both streams are then ranked by the same model, which predicts a probability for each action you might take and blends them with weights held in the code. Whether a post can be shown at all is decided separately, by visibility filtering, from your own blocks and mutes and from labels other systems attach.

03·NVIDIA

PersonaPlex is a fine-tune of Moshi, not a new architecture

The README says so plainly: it is based on the Moshi architecture and weights, benefiting from the Helium model underneath. That is not a criticism — it is why the full-duplex behaviour works — but it means the thing to read before betting on it is Moshi’s literature, and it explains why the server command is moshi.server rather than something NVIDIA-branded.

04·Anthropic

Six plugins, not four, and one of them is load-bearing

Alongside the pitch agent, GL reconciler and market researcher there are investment-banking and equity-research plugins. More importantly financial-analysis is the core: it carries the shared modelling skills and all eleven data connectors the others rely on, so it goes on first rather than as an afterthought.

05·Cloudflare

Gatekeepers solve the approval problem, which is the actual idea

Described as guardrails, they are more specific than that. When an agent does something needing approval, the Gatekeeper simulates the result locally and tells the agent it succeeded, so the agent keeps queueing work instead of blocking. You approve or reject the queue later, in bulk. That is a direct answer to the choice between babysitting every action and giving up and passing --dangerously-skip-permissions. Worth knowing the repo is version 2, a complete rewrite, and Cloudflare calls the local mode a way to try it rather than to run on.

06·Tencent

The hardware floor is exact, the speed claim is not

The README states 6 GB of VRAM for shape generation and 16 GB for shape plus texture — those numbers are verbatim, and they are the reason texture work needs a serious card. The widely quoted sub-second figure does not appear anywhere in the repository, so treat it as marketing until someone shows a benchmark; what the repo does document is the Turbo and FlashVDM distilled variants as the fast path.

07·Alibaba

Ten languages, and a three-second clone

The ten are Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish and Italian, listed per model rather than in aggregate. The base models do voice cloning from roughly three seconds of audio, the custom-voice models ship nine premium timbres, and everything streams. Sizes are 1.7B and 0.6B, so this is the lightest thing on the list.

NVIDIA

PersonaPlex — the demo server

Needs an NVIDIA GPU and Linux. Accept the licence on the model page before the token will work.

personaplex.sh11 lines
# Ubuntu / Debian — the README also covers Fedora and RHEL
sudo apt install libopus-dev

git clone https://github.com/NVIDIA/personaplex.git
cd personaplex
pip install moshi/.

# accept the model licence on the Hugging Face page first,
# then use a token from that account
export HF_TOKEN=YOUR_HUGGINGFACE_TOKEN
SSL_DIR=$(mktemp -d); python -m moshi.server --ssl "$SSL_DIR"
Cloudflare

Cloudflare OS — locally

cloudflare-os.sh3 lines
git clone https://github.com/cloudflare/cloudflare-os.git
cd cloudflare-os
pnpm run-local          # then open http://localhost:8787
Anthropic

The finance plugins

financial-services.sh13 lines
claude plugin marketplace add anthropics/financial-services

# the core plugin: shared modelling skills, and all 11 data
# connectors that the rest depend on. Install it first.
claude plugin install financial-analysis@claude-for-financial-services

claude plugin install pitch-agent@claude-for-financial-services
claude plugin install gl-reconciler@claude-for-financial-services
claude plugin install market-researcher@claude-for-financial-services

# two the guide leaves out
claude plugin install investment-banking@claude-for-financial-services
claude plugin install equity-research@claude-for-financial-services
Tencent

Hunyuan3D-2 — photo to model

hunyuan3d.sh14 lines
git clone https://github.com/Tencent-Hunyuan/Hunyuan3D-2.git
cd Hunyuan3D-2
pip install -r requirements.txt
pip install -e .

# textures only — skip both if you just want geometry
cd hy3dgen/texgen/custom_rasterizer && python3 setup.py install && cd ../../..
cd hy3dgen/texgen/differentiable_renderer && python3 setup.py install && cd ../../..

# the Turbo path
python3 gradio_app.py --model_path tencent/Hunyuan3D-2 \
  --subfolder hunyuan3d-dit-v2-0-turbo \
  --texgen_model_path tencent/Hunyuan3D-2 \
  --low_vram_mode --enable_flashvdm
Alibaba

Qwen3-TTS

qwen-tts.sh4 lines
pip install -U qwen-tts

# the README asks for a fresh, isolated environment and suggests
# Python 3.12; the package metadata itself allows 3.9 and up.
The commercial angle

What the source suggests charging

These are the write-up’s own suggested ranges, not observed market rates. Included because the licence question only matters if you are billing for it.

ReleaseThe serviceSuggested range
SymphonyAgent-run development retainer, priced by tasks cleared$2,000–6,000 / month
x-algorithmX growth strategy and writing built on the real signals$1,000–3,000 / month
PersonaPlexAI phone receptionist for businesses that miss calls$1,000–3,000 setup, $300–800 / month
financial-servicesOutsourced analyst: monthly reporting and reconciliations$500–2,000 / month
Cloudflare OSAI workspace set up around a team’s own tools$2,000–5,000 setup, then monthly
Hunyuan3D-23D product models for online stores$50–150 per model
Qwen3-TTSVoiceover and multi-language dubbing$50–300 per video
Before you bill for any of it

The checks that decide whether this is a business

Where the client is — Tencent’s licence excludes the EU, the UK and South Korea outrightHow big the client gets — past one million monthly users, Hunyuan3D needs a separate Tencent licenceWhether you accepted the model licence, not just cloned the repo — PersonaPlex gates its weightsWhether callers are told they are talking to a machine, which many jurisdictions requireWhether you have written consent for any voice you clone, the client’s own voice includedWho signs off the finance output — these agents draft, a qualified person is accountableThat previews are previews: Symphony and Cloudflare OS both say so in their own READMEs