Bittensor subnet news
Subnet teams' own channels, X posts, press, newsletters and events read from the chain — GitHub lives in its own tab
SOMA@SomaSubnetSOMA’s Claude Code competition is underway. Are you in? thesoma.ai/dashboard
Linkthesoma.aithesoma.ai/dashboardSOMA@SomaSubnetSOMA now plugs into your @OpenCode subscription. Try it with 30 days free trial: thesoma.ai Build more.
Linkthesoma.aithesoma.aiSN114 large sell: 426.8 τ
Taosis newsletters
Taosis News: the week’s articles from the subnet teams, the press and community newsletters, every Friday. Taosis Weekly: the week’s movers, best yields and chain events, with dollar values, every Monday. Past issues
Validator Arbos (5GTW…ySXP) scheduled a child-key change on 6 subnets
Validator hotkey 5GTWwu3NCbYmVFzwfrXa1sx8y4UNLR3padWfVZwjNnDmySXP announced new child keys on SN112, SN114, SN117, SN120, SN124, SN128, applying at block 9,435,741 (~24h). Stake delegated to it there routes through its children after that.
ValidatorArbosSubnetsSN112, SN114, SN117, SN120, SN124, SN128
Between blocks 9,428,540 and 9,428,599
SOMA@SomaSubnetKimi K3 is live on SOMA. Use it in Claude Code, Codex, or GitHub Copilot CLI, with context compressed before every request hits the model. Pick Kimi K3 in the app and start: thesoma.ai/
Linkthesoma.aithesoma.ai/SOMA@SomaSubnet@LeadpoetAI Let's talk enterprise.thesoma.ai/
Linkenterprise.thesoma.aienterprise.thesoma.ai/SOMA@SomaSubnetWe're partnering with Leadpoet to bring SOMA to more enterprise teams running AI in production. @LeadpoetAI will help us find and reach the companies, and the people who make the calls there, that could benefit from context compression.
On Taosis
Swap TAO ⇄ USDC
USDC on Ethereum to TAO, or back. No KYC, wallet to wallet; the fee and the amount you receive shown before you send. Not instant: run by Taosis, paid by hand within hours.
SOMA@SomaSubnetSave 20% or more on your enterprise inference. How? SOMA compresses your prompts before they reach the model. Fewer tokens, same results. enterprise.thesoma.ai/
Linkenterprise.thesoma.aienterprise.thesoma.ai/SOMA@SomaSubnetGLM 5.3 and GLM 5.3 Flash from @Zai_org are live on SOMA. Our latest integration brings new models to SOMA enabling you to save tokens. Connect them here: thesoma.ai Compress. Pay less.
Linkthesoma.aithesoma.aiSOMA@SomaSubnetSOMA Highlights - October 26 Claude Code support, three new models (Le Chonk, Kimi K3, GLM 5.3), token savings up from 30% to 40% after the first Jev round, and our first closed-source model: GPT-6 Luna. Compress. Pay less.
SOMA@SomaSubnetWe believe ALL agents can run faster (thanks to SOMA).
Quoting @615818451
$100k to the developer who can make muse spark open fastest! twitter.com/danielgross/status/2113351902402613…
SOMA@SomaSubnetNew competition - time for GPT 6 Luna! Our first round with closed-source model. This opens up new opportunities for SOMA and brings us one step closer to supporting compression across the entire AI ecosystem.
SOMA@SomaSubnetFrom 30% to 40% token savings after the first competition with Jev! A 10pp jump in just two weeks. Mind-blowing results.
SOMA@SomaSubnetAll for one, one for all. That was great to launch three new Bittensor integrations in the past month: @say_gm_, @engyai, @chutes_ai. Try SOMA: app.thesoma.ai/register
Linkapp.thesoma.aiapp.thesoma.ai/registerSOMA@SomaSubnetClaude Code is now supported on SOMA. Our latest harness integration joins Codex and Copilot, bringing SOMA to more of the tools you code with. Try Claude Code with SOMA: app.thesoma.ai/login
Linkapp.thesoma.aiapp.thesoma.ai/loginSN114 large sell: 457.3 τ
SN114 large buy: 454.8 τ
SOMA@SomaSubnetMeanwhile, in our datacenter... You can fit more into your @MistralAI context window and let it work longer for the same price and quality. Now available for Mistral Large 4. Register: app.thesoma.ai/register
Linkapp.thesoma.aiapp.thesoma.ai/registerValidator Astrid (5ELR…qwVh) scheduled a child-key change on 20 subnets
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.
ValidatorAstridSubnetsSN6, SN8, SN9, SN20, SN21, SN44, SN47, SN53, SN54, SN61, SN65, SN73, … (20 subnets)
Between blocks 9,285,716 and 9,285,779
SOMA SN114 puts a CPU budget around agent tests
Bittensor's SOMA SN114 changes how validators run agent tests, limiting resources and the CPU count shown to software that creates parallel workers.
Tao Outsidersource ↗
SOMA@SomaSubnetWe have the best context compressor around! 🧵 1/4 SOMA can provide 30% more effective compute to your system. This gain has been measured against an unchanged baseline, with identical tasks, model, provider, environment and generation settings. thesoma.ai/labs/experiment-methodology
Linkthesoma.aithesoma.ai/labs/experiment-methodologySOMA@SomaSubnet🧵 4/4 We didn't stop at comparing SOMA against the baseline. We also benchmarked SOMA head-to-head against Caveman on a separate test set, using DeepSeek V4.1 Flash - the same model currently used in our competition. The result? SOMA delivers greater token savings AND a higher benchmark score.
SOMA@SomaSubnet🧵 3/4 We've benchmarked SOMA across 6 models, with token savings ranging from 18.91% to 47.10%, while maintaining comparable or better solve rates. We repeat runs because agent trajectories are stochastic: one run can be short while another loops, explores a dead end or hits provider noise. Repeating runs lets us distinguish stable improvements from lucky runs.
SOMA@SomaSubnet🧵 2/4 The basic truth is that every unnecessary token sent to a model consumes inference capacity. And so, removing unnecessary context translates into: - lower cost - lower model workload - higher throughput from the same hardware
SOMA@SomaSubnetWe have the best context compressor around! 🧵 1/4 SOMA can provide 30% more effective compute to your system. This gain has been measured against an unchanged baseline, with identical tasks, model, provider, environment and generation settings. Check out the details: thesoma.ai/labs/experiment-methodology
Linkthesoma.aithesoma.ai/labs/experiment-methodologyValidator TAO.app (5CoZ…bifj) scheduled a child-key change on 125 subnets
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.
ValidatorTAO.appSubnetsSN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8, SN9, SN10, SN11, SN12, … (125 subnets)
Between blocks 9,274,365 and 9,274,433
SN114 large sell: 470.9 τ
SN114 large buy: 470.7 τ
Validator Tensorplex Labs (5E4z…ULde) scheduled a child-key change on 4 subnets
Validator hotkey 5E4z3h9yVhmQyCFWNbY9BPpwhx4xFiPwq3eeqmBgVF6KULde announced new child keys on SN102, SN109, SN114, SN121, applying at block 9,279,905 (~24h). Stake delegated to it there routes through its children after that.
ValidatorTensorplex LabsSubnetsSN102, SN109, SN114, SN121
Between blocks 9,272,702 and 9,272,768