Subnet teams' own channels, X posts, press, newsletters and events read from the chain — GitHub lives in its own tab
TPN (SN65) cuts an 8B AI model's size by 57% while improving its knowledge score in TPN-005, then launches TPN-006 with a tighter 8GiB memory limit for AI on everyday devices. [...] The post TPN-005 Reduced Model Size by 57%, TPN-006 Pushes For Tighter Memory Limits appeared first on The TAO Daily.
The TAO Dailysource ↗
True Performance Network@TPN_LabsNext month is going to be a big one for TPN 👀 We’ve been building something that we want to share with the world. Powered by TPN_Labs - Subnet 65 $TAO
Validator hotkey 5HK2LZRrvj581VjyxuFxb6ehFZoNi5PqAg3Y2HsmYPhGYYXr announced new child keys on SN31, SN65, SN118, applying at block 9,301,087 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,293,869 and 9,293,932
Validator hotkey 5HK1Ubcg6MBwDGW7Td4RtAvE1FNyvkHvXKfwymW9q1BCRJAK announced new child keys on SN31, SN65, SN118, applying at block 9,301,084 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,293,869 and 9,293,932
Validator hotkey 5F7dvCS7mdzJU4o9bAUx4LPgpF3XeMM1XEz9nVqVNga9r6LW announced new child keys on SN31, SN65, SN118, applying at block 9,301,090 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,293,869 and 9,293,932
Validator hotkey 5EqYdsxW1EYfb7PdJrazCYd1JuunqhABtHSWKh34jPbQoaUf announced new child keys on SN31, SN65, SN118, applying at block 9,301,093 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,293,869 and 9,293,932
True Performance Network@TPN_LabsTPN-006 is live. TPN-005 produced a model that runs on consumer hardware; this round asks whether it can also run on weaker hardware. Competition 6 will make Ministral-3-8B Uncensored even smaller, lowering the size cap to 8GiB (down from 11GiB in TPN-005) while still holding 32k context. One winner takes the full 25% emission weight, King of the Hill style, scored on Fluid_knowledge. Every GiB down means more devices can run the model, bringing local and private AI to more people.
Validator hotkey 5DjtmV76CRvxZqd68fDwWyTmsfqUZ5EUXVJ21FKTBzwLT4Kd announced new child keys on SN39, SN50, SN52, SN55, SN61, SN62, SN65, applying at block 9,294,183 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,286,982 and 9,287,045
Validator hotkey 5ELREhApbCahM7FyGLM1V9WDsnnjCRmMCJTmtQD51oAEqwVh announced new child keys on SN6, SN8, SN9, SN20, SN21, SN44, SN47, SN53, SN54, SN61, SN65, SN73 and 8 more, applying at block 9,292,933 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,285,716 and 9,285,779
Validator hotkey 5HK2LZRrvj581VjyxuFxb6ehFZoNi5PqAg3Y2HsmYPhGYYXr announced new child keys on SN65, applying at block 9,290,287 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,283,050 and 9,283,115View subnet →
Validator hotkey 5HK1Ubcg6MBwDGW7Td4RtAvE1FNyvkHvXKfwymW9q1BCRJAK announced new child keys on SN65, applying at block 9,290,284 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,283,050 and 9,283,115View subnet →
Validator hotkey 5F7dvCS7mdzJU4o9bAUx4LPgpF3XeMM1XEz9nVqVNga9r6LW announced new child keys on SN65, applying at block 9,290,290 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,283,050 and 9,283,115View subnet →
Validator hotkey 5EqYdsxW1EYfb7PdJrazCYd1JuunqhABtHSWKh34jPbQoaUf announced new child keys on SN65, applying at block 9,290,293 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,283,050 and 9,283,115View subnet →
True Performance Network@TPN_LabsHere’s the case study for the TPN-005 model. We asked the network for a purpose built LLM that could fit inside a chat interface’s memory and perform well on general knowledge. The subnet produced a successful output, a model optimized for general knowledge within a fixed memory budget, performing well on the requested capability. Every figure in the study links to an immutable run record, from the model revision, sealed epoch, to the evaluation settings, so everything is verifiable.
Quoting @TPN_Labs
We asked for a model that fits on a laptop and knows things. A case study in purpose-built model optimization. x.com/i/article/2110049958774603776
True Performance Network@TPN_LabsWe asked for a model that fits on a laptop and knows things. A case study in purpose-built model optimization. x.com/i/article/2110049958774603776
Validator hotkey 5CoZxgtfhcJKX2HmkwnsN18KbaT9aih9eF3b6qVPTgAUbifj announced new child keys on SN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8, SN9, SN10, SN11, SN12 and 113 more, applying at block 9,281,605 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,274,365 and 9,274,433
True Performance Network@TPN_LabsA great report on how TPN v2.0 increased the subnet's quality of output.
Quoting @taomedia_
✏️ Story by @thillerich12 Read about @TPN_Labs on our website: tao.media/tpn-labs-releases-first-production-mo…
True Performance Network@TPN_LabsA great 4 min read on TPN005 and upcoming plans. Shout out to @taodaily_io for always being on top of any news Bittensor has to offer. Great team, awesome work!
Quoting @taodaily_io
What if you could make an AI model 40% smaller without taking away its ability to handle long conversations and documents? That is what @TPN_Labs (Subnet 65 on Bittensor) just achieved with TPN005. The model keeps Ministral-3-8B’s 32K context window while using a much smaller footprint. Full story👇 taodaily.io/tpn-just-made-an-ai-model-40-smalle…
True Performance Network@TPN_LabsWe're glad to see that many people in the Bittensor community are starting to understand the vision. Thank you @vaNlabs and the Bittensor community, it's this type of support that accompanies teams in achieving great results 🙏
Quoting @vaNlabs
This is what I like to see from subnets. Miners produce something real and it ships in a product. If TPN's consumer app runs on models the Bittensor network made, the subnet is doing the work, not just the marketing | $TAO twitter.com/TPN_Labs/status/2109627952514175330
Validator hotkey 5HK2LZRrvj581VjyxuFxb6ehFZoNi5PqAg3Y2HsmYPhGYYXr announced new child keys on SN65, applying at block 9,279,487 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,272,245 and 9,272,309View subnet →
Validator hotkey 5HK1Ubcg6MBwDGW7Td4RtAvE1FNyvkHvXKfwymW9q1BCRJAK announced new child keys on SN65, applying at block 9,279,484 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,272,245 and 9,272,309View subnet →
Validator hotkey 5F7dvCS7mdzJU4o9bAUx4LPgpF3XeMM1XEz9nVqVNga9r6LW announced new child keys on SN65, applying at block 9,279,490 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,272,245 and 9,272,309View subnet →
Validator hotkey 5EqYdsxW1EYfb7PdJrazCYd1JuunqhABtHSWKh34jPbQoaUf announced new child keys on SN65, applying at block 9,279,493 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,272,245 and 9,272,309View subnet →
Validator hotkey 5ELREhApbCahM7FyGLM1V9WDsnnjCRmMCJTmtQD51oAEqwVh announced new child keys on SN8, SN9, SN12, SN19, SN20, SN21, SN23, SN25, SN41, SN44, SN47, SN52 and 17 more, applying at block 9,278,533 (~24h). Stake delegated to it there routes through its children after that.
Between blocks 9,271,302 and 9,271,367
TPN (SN65) unveils TPN005, its first production-ready AI model, 40% smaller than Ministral-3-8B while preserving its 32K context window through Bittensor competition. [...] The post TPN Just Made an AI Model 40% Smaller, and This Could Be a Big Deal appeared first on The TAO Daily.
The TAO Dailysource ↗
The first TPN v0.2 winner compresses Ministral-3-8B by about 40% while retaining its full 32,000-token context window.
Intelligence — tao.mediasource ↗
True Performance Network@TPN_LabsHuge thank you to @ActuallyMentor, the best CTO that TPN Labs could ask for. We build in public.
Quoting @ActuallyMentor
It's been so fun working on the improved versions of this. Super cool learning curve for the team as well on some avant-garde techniques. Let's just say that future competitions are going to go waaaaay beyond simple benchmarking. twitter.com/TPN_Labs/status/2109627952514175330
True Performance Network@TPN_LabsThank you @const_reborn We hope you'll enjoy the upcoming developments, as there is a lot yet to announce as well as a lot of progress to look forward to. TPN is Bittensor and Bittensor is TPN. $TAO
Quoting @const_reborn
Bravo to the newest neo lab on Bittensor twitter.com/tpn_labs/status/2109627952514175330