Chutes, the largest inference subnet on the Bittensor network, has stopped selling its monthly subscription plans and moved to pay-as-you-go credits, with its two newest beta models, Qwen3.8 Max and DeepSeek V4.1 Pro, reachable only on usage billing. The change resets the commitment math on a major decentralized network — and it comes with a 35% discount on two flagship models that the operator says expires November 23.
Key facts
- Chutes' pricing page now states: "Monthly plans are no longer sold on chutes.ai" — existing plans "stay active until the end of your current billing period and then do not renew."
- The change was announced by the Chutes team on October 26, as reported by Tao.media: "subscriptions are no longer feasible" due to the mechanics of the new model offerings and "current general GPU/compute market dynamics."
- Tao.media reports the two new beta models are excluded from subscriptions and are PayGo only, per the operator's announcement.
- The same report says neither new model currently supports Chutes' TEE, confidential-compute, or end-to-end encryption modes.
- On Chutes' live pricing table, the comparable flagship tier runs $2.00 input / $6.00 output per million tokens (Qwen3.8-2.4T-A95B and Qwen3.8-Max-0902), with per-model rates for 19 listed models.
- The subscription tiers being retired were priced at Base $3, Plus $10, and Pro $20 per month (checked November 7, 2026).
What happened
Chutes is Bittensor subnet 64, a serverless inference platform where GPU operators serve open-weight models through an OpenAI-compatible API. Until late October it ran a hybrid billing model: three monthly subscription tiers ($3, $10, and $20 with daily request quotas and small overage discounts, per Activepieces' November pricing check) sat alongside per-token pay-as-you-go.
That ended on October 26, when Tao.media reports the Chutes team announced Qwen3.8 Max and DeepSeek V4.1 Pro in beta and, in the same platform update, the end of subscriptions. The quoted reasoning: "due to the mechanics of these new model offerings and current general GPU/compute market dynamics, subscriptions are no longer feasible."
Two details in the announcement matter more than the headline. First, the new models are not part of any legacy plan — a subscriber who wants them pays usage rates on top of a plan that is already running out its clock. Second, neither model supports the platform's TEE, confidential-compute, or end-to-end encryption modes right now. Chutes markets itself heavily on confidential compute (its pricing page marks "TEE only" models), so a flagship tier without that profile is a real difference for anyone routing sensitive workloads.
Tao.media's coverage matches the announcement on both points, and the operator's pricing page has since been updated to codify the new state: billing is pay-as-you-go "with no subscription and no minimum commitment," monthly plans are no longer sold, and there is no free tier. The current public model table lists 19 models with per-token rates — for example, Qwen3.8-Max-0902 and Qwen3.8-2.4T-A95B at $2.00 per million input tokens and $6.00 per million output tokens, DeepSeek-V4.1-Flash-1017 at $0.195/$0.78, and GLM-5.3 at $0.80/$2.50.
One caveat on sourcing: the headline numbers here are operator-published. The billing-structure change is confirmed by two surfaces under the operator's own control (the announcement and the pricing page) plus independent coverage, but a reproducible probe of the renewal flow — the verification bar this story ultimately needs — has not been published, and the 35% discount on GLM-5.3 and DeepSeek V4.1 Flash that folds into this story (a Chutes promotion under an Alibaba Cloud model-access arrangement, expiring November 23) is a provider claim until it shows up on an invoice or an independent check.
Why it matters
The flat-rate tier was the lowest-commitment way onto a decentralized network: $3 got a month of quota-backed requests with no forecasting. Losing it pushes every builder into the same posture as a conventional API account — estimate token volumes, manage a prepaid balance, eat the variance. For occasional users, that is a worse deal in expectation; for heavy users, it removes the subsidy they were getting under quota pricing. DeAI's comparison coverage has long noted that Chutes' subscription tiers were the outlier in a market that bills per token; that outlier is now gone.
The timing matters too. DeAI's tracker cycle 8 recorded the network's own admission of a 45% six-week volume drop last month, and our margin-claims coverage tracked how subscription economics strain GPU-backed inference when compute prices rise. The operator's stated reason — that flat-rate plans are "no longer feasible" under current GPU market dynamics — reads as the pricing structure catching up with the input-cost reality, not a pivot away from Bittensor: the pricing page still advertises payment in TAO and Bittensor alpha tokens.
For builders deciding where to run models this week, the practical decision points are the November 23 discount expiry on two flagship models and the security profile of the new tier: if your workload needs confidential compute, the PayGo-only beta models are currently the wrong destination, and the TEE-marked models on the pricing table are where that capability lives.
Background
Chutes built its user base on Bittensor's incentive mechanism: validators direct TAO emissions to GPU operators serving models, which lets the platform undercut conventional cloud pricing for open weights. DeAI's explainer on decentralized inference networks covers how that subsidy model works and where it strains. The subscription tiers were an attempt to package that cheap supply for non-developers — a convenience layer most per-token providers never offered.
That layer survived its first compute-price shock in October, when the network acknowledged the volume drop. Cycle 8 of the provider tracker also showed the verification gap: nine direct probes of Chutes' endpoint returned rate limits, and the ledger still reads zero independently verified field values across 17 tracked providers. A subscription-to-paygo cutover is exactly the kind of operational claim that a renewal-flow probe or a ToS diff could grade — nobody has run one yet.
What's next
The 35% discount on GLM-5.3 and DeepSeek V4.1 Flash expires November 23 — after that, the operator's pricing table is the only rate that matters. Watch for Qwen3.8 Max and DeepSeek V4.1 Pro rates to appear on the public pricing table as the models leave beta, and for TEE support on the new tier. DeAI's weekly check-harness runs will probe both.
Questions
- Did Chutes end its subscription plans?
- Chutes says it has stopped selling monthly subscriptions and that existing plans will not renew at the end of their current billing period. Its pricing page now states: "Monthly plans are no longer sold on chutes.ai," and billing there is described as pay-as-you-go with no subscription and no minimum commitment.
- Which Chutes models are pay-as-you-go only?
- Qwen3.8 Max (Qwen/Qwen3.8-Max-0902) and DeepSeek V4.1 Pro (deepseek-ai/DeepSeek-V4-Pro-0813) in beta. The operator says both are excluded from legacy subscriptions and PayGo-only. Neither model appears on Chutes' current public pricing table, so published rates for them are not yet listed.
- Do Chutes' new beta models support confidential computing?
- No, according to the operator's October 26 announcement: Qwen3.8 Max and DeepSeek V4.1 Pro beta do not currently support Chutes' TEE, confidential-compute, or end-to-end encryption modes. Builders who need those protections are pointed to the models that run on Chutes' marked TEE compute.
- Is Chutes leaving the Bittensor network?
- No evidence of that. Chutes is Bittensor subnet 64, and its pricing page continues to advertise payment in TAO and Bittensor alpha tokens alongside Stripe methods. The billing change affects subscription plans on chutes.ai, not the subnet's role in Bittensor's incentive network.
Sources
About DeAI
DeAI is an independent publication covering open-weight AI models, private inference, and decentralized infrastructure — the tools for running AI you actually control. We test providers on price, privacy, and refusal behavior and publish the numbers, not the vibes. DeAI is powered by Morpheus (mor.org), a decentralized inference marketplace, and covers it on the same terms as every other provider.
Powered by Morpheus and StrandCMS
Morpheus is a decentralized inference marketplace, covered on the same terms as every other provider — we rank it wherever the data lands. StrandCMS is the open-source, agent-first framework this site is built on.
