Bittensor subnet news

True Performance Network@TPN_Labs

Great steps for @Gratis_AI and the team. The work must go on, if you thought loading models for free was the endgame, we'd like to show you what's coming next. This November, gratis-ai.com on @ProductHunt

Quoting @Gratis_AI

This month, @gratis_AI will launch on @ProductHunt We hope to see you all there to test gratis and give us feedback. If you want to be part of the launch, make a Product Hunt account now, so you won't have to later!

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

TPN's 8th model is under production! We shipped 7 optimized models on @huggingface, coming from different labs like @googlegemma, @Alibaba_Qwen and @MistralAI. Competition n8 is the largest one that TPN has ever hosted, with the scope to take @Google's Gemma-4-31B under a 25GiB ceiling. The rewards for the winner have increased, as the task at hand became more complex. We wish good luck to the participants, and may the best model surface as winner.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Another stunning @Gratis_AI product feature clip from @hammertime_one 💪 gratis-ai.com

Quoting @hammertime_one

Anthropic, OpenAI and Google run AI in the cloud. Gratis runs it on your laptop Cloud AI works like a restaurant. You order, someone cooks it far away, the answer arrives. You never see the kitchen, and the kitchen sees everything you ordered. On Nov 2, @TPN_Labs (Bittensor SN65) launched Gratis to hand you the kitchen. Open source AI models run inside a browser tab. Or a desktop app, if you want more power. Free. No setup. No signup. Pick a model, or load any model from Hugging Face. It downloads to your machine and runs there. No account. No credits. No cloud. Every chat stays private by design. TPN builds small, capable models. Gratis runs them. The big labs rent you intelligence by the prompt. This one you just open.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Try gratis-ai.com, share the model you choose and enter the $100 giveaway! Curious to know which models are working best for people! @Gratis_AI

Quoting @Gratis_AI

$100 giveaway 🎉 We want to celebrate the launch of gratis-ai.com with a giveaway! $20 to 5 people! Rules: 1⃣ Follow @Gratis_AI and @TPN_Labs 2⃣ Share a screenshot of the model you loaded on gratis-ai.com 3⃣♥️ & 🔄 this post 4⃣@ tag 2 frens in the comments Winners will be announced below this post in one week.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Over 2600 unique visitors and 160k+ total requests on @Gratis_AI in the past 48 hours. Did you try it yet? It's free🐰

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

@huggingface @Gratis_AI TPN-007 model huggingface.co/tpnlabs/tpn-007-base

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

It's free 🐰gratis-ai.com

Quoting @YakusaHL

Big update from @TPN_Labs TPN just introduced Gratis, a free way to run open-source AI models directly on your own device. No complicated setup. No need to send every prompt to a cloud server. You can load models from @huggingface and run them locally through your browser or desktop app, keeping your conversations private by design. What I like about this launch is how it connects with TPN's core mission. TPN works on making AI models smaller and more efficient. Gratis brings that work directly to everyday users. From model optimisation to a product people can actually use. Another interesting development from Bittensor SN65. $TAO 🧠 Try it: gratis-ai.com

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Awesome @Gratis_AI features highlight. Very cool work @hammertime_one🐰

Quoting @hammertime_one

ChatGPT knows more about you than your best friend does. Gratis couldn't read your chats if it tried why? the model runs on your machine @TPN_Labs launched Gratis on Nov 2. Open source AI inside a browser tab, or a desktop app for more power free, no setup, no signup pick a model, or load any model from Hugging Face it downloads, it runs, locally no account, no credits, no cloud TPN builds small, capable models on Bittensor SN65 Gratis runs them one click, zero servers

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

TPN-007 is now live on @huggingface The task was to fit Qwen3.5-4B under a 4.2 GiB ceiling without losing its reasoning, instruction following or multilingual ability. The produced model stands at 2.32 GB and it runs on @Gratis_AI 🐰 Yet another milestone for subnet 65!

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

TPN x Shizzy Unchained🎙️ @0xDutchmikel discusses TPN subnet history, the product line, and building in the Bittensor ecosystem from the perspective of a subnet founder. Thanks to @ShizzyUnchained for the great interview!

Quoting @ShizzyUnchained

AI on Your Terms? | True Performance Network | Bittensor Subnet 65 x.com/i/broadcasts/2qVAkEARXAVJVX

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Nice visual representation of subnet 65 main scope of work!

Quoting @taoleeh

Bittensor SN65 is a contest to make models smaller, and the winners just got a front door. Each round the network names a base model and a memory ceiling. Miners send in shrunken versions that still have to clear fixed benchmark scores, and validators rerun every score themselves along with a KL divergence check (how far the small model's next-token guesses drift from the original). Numbers you can't reproduce get thrown out before ranking. @TPN_Labs turned that output into Gratis. It reads your RAM, tells you which models fit with room to spare, downloads one, then runs offline on your own chip. Miners supply the models, the app puts them in front of people who will never open a terminal. Most subnets stop at a Hugging Face repo and call it shipped. A file sitting in a repo is inventory. A download button is a product. What else in the AI stack gets better once strangers are paid to beat a fixed target? x.com/TPN_Labs/status/2117255539235180638 $TAO #Bittensor

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

gratis-ai.com

Quoting @i_compounding

This has been the easiest deployment of this kind of tech I've ever experienced.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

TPN Labs episode with @ShizzyUnchained and @0xDutchmikel coming tomorrow🐰

Quoting @ShizzyUnchained

The AI you want. The hardware you actually own. Can TPN bridge the gap? Tomorrow I’m live with Mikel from @TPN_Labs, Bittensor SN65, to unpack model compression, benchmark tradeoffs and running AI yourself. Tuesday, November 3, 2026 11:30 AM ET / 8:30 AM PT Set a reminder and bring your questions: youtube.com/watch?v=1ajAfGr_dh0

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Local models are capable of resolving everyday tasks. Most often, there is no need to set frontier LLMs to max effort. Run your tasks through gratis-ai.com instead.

Quoting @ActuallyMentor

@TPN_Labs @Gratis_AI @huggingface I keep being surprised at how well in-browser models work on Macbooks. And that's without installing the native apps.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

We're excited to introduce @Gratis_AI. Run opensource AI models locally on your machine for free, with no setup required. Gratis runs models from @huggingface directly in a browser tab or desktop app for more power, making every chat private by design. TPN builds small, capable models; gratis runs them. gratis-ai.com

Quoted post

Introducing gratis-ai.com Run open source AI models locally in your browser, in just a few clicks. Pick an opensource model or load any model from @huggingface. It downloads to your machine and runs there privately; every prompt stays on your device. Get your free personal AI today.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Small devices will require small purpose built models.

Quoting @VenturaLabs

Does every AI job need a frontier model, or is most of the money going to brains way bigger than the task?

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

@Alibaba_Qwen Competition Dashboard: stats.trueperformancenetwork.com/competition?id…

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

TPN-007 is live. The network has been asked to fit @Alibaba_Qwen 3.5-4B under a 4.2GiB ceiling, without losing its reasoning, instruction following or multilingual ability. 007 the first competition running on the revised engine, with new KL divergence and quantization logic, wider architecture support, and a fresh sealed epoch of fluid_knowledge. Small, purpose built models are what will run on local machines.

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

New: Validate a quant tool. With TPN Bench quant verification tool, users can deterministically or probabilistically prove that a quantization is indeed a quant of the base model. The tool's report shows how often the quant picks the same next word (top-1 match), how far its predictions drift (KLD, KL divergence) and how much worse it predicts text (perplexity) compared to the parent model. This is something that we will definitely use internally, but it is also a tool that devs and labs might use if they have a quantization, in order to verify their quants independently. Now live on bench.trueperformancenetwork.com

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

What is TPN Labs building? Join @ActuallyMentor and @VenturaLabs to delve deep into what TPN Labs is building through subnet 65. Special thanks to Grant, the best host 🙏 $TAO

Quoting @VenturaLabs

Ep. 106 - Mentor Palokaj Mentor Palokaj @ActuallyMentor is building TPN @TPN_Labs Timestamps 1:01 - From a TAO Bridge to a VPN Subnet 4:28 - The 440 Upgrade and a Six-Hour Rethink 6:42 - Profitable, but Not Parabolic 9:42 - What TPN Actually Does to a Model 10:03 - Quantization, Explained With a Ruler 11:36 - Quantization-Aware Distillation 12:25 - Nobody Knows How to Measure an LLM 13:15 - Anyone Who Can Cheat Will Cheat 13:49 - Fluid Bench, a Benchmark That Moves 15:13 - The TPN Workbench and Optimize 16:57 - An LLM as a Plinko Machine 20:02 - Toward Models That Are Experts 21:12 - DeepSeek on 8GB of VRAM? 21:30 - The Outlier Problem 25:18 - What Each Halving Really Costs 26:18 - Ternary Weights and Hadamard Rotations 32:09 - Is There an Optimal Bit Rate? 34:05 - Where TPN Goes Next 35:03 - Treadmills and a Teaser

Full post on TaosisOpen on X ↗

True Performance Network@TPN_Labs

Compare is live on Bench: bench.trueperformancenetwork.com/

Full post on TaosisOpen on X ↗