Skip to content
inetGeek

Together AI pricing and limits

iScore47–64 · ≈#6/10
www.together.ai · LLM APIs
SEPT 2026 audit

Serverless per-token inference for open-weight models from other labs, plus dedicated GPU endpoints, hourly GPU clusters and fine-tuning, paid from prepaid credits.

Visit Together AI documentation

No affiliate links, sponsored placements or paid rankings appear on this site. Ordering follows the sourced data and the stated criteria.

Entry paid planNo plan fee, but the platform is fully prepaid: access requires a minimum $5 credit purchase, and API access is suspended when the balance reaches zero. Purchased credits do not currently expire.
RegionsServerless requests are routed across Together's own infrastructure with no region selection; DeepSeek models are named as hosted in North America data centers. Region choice, including EU regions, comes through dedicated endpoints or an arrangement with Together.
Free tierNo
CategoryLLM APIs

Reviewed by Palash BagchiUpdated 22 criteria comparedSources last checked

Together AI pricing at a glance

Entry paid plan
No plan fee, but the platform is fully prepaid: access requires a minimum $5 credit purchase, and API access is suspended when the balance… source ↗
Pricing model
Four products priced separately: serverless inference per 1M input and output tokens for each model, with a lower cached-input rate on… source ↗

First line of each sourced value. The full text is in Together AI pricing below.

Together AI prices by model

Serverless API prices per modelServerless chat-model prices in USD per 1M tokens as published on Together AI's pricing page, in the order the page lists them. Notes gives the cached-input rate the page prints beside the input price, where there is one.
ModelInput (per 1M tokens)Output (per 1M tokens)Notes
MiniMax M3$0.30$1.20Cached input $0.06.
Kimi K3$3.00$15.00Cached input $0.30.
Qwen3.8-2.4T-A95B$2.00$6.00Cached rate not shown: Together's pricing page and docs disagree.
Muse Glimmer 30B$0.35$1.50Cached input $0.04.
DeepSeek V4 Pro 0813$1.32$3.96Cached input $0.13.
GLM-5.3$1.40$4.40Cached input $0.26.
GLM-5.3-Flash$0.15$0.50Cached input $0.03.
DeepSeek V4.1 Flash$0.30$1.20Cached input $0.006.
Gemma 4 31B$0.39$0.97No cached rate listed.
Qwen3.7-Plus$0.32$1.28No cached rate listed.

Pricing | Together AI · read · every row quoted from this page

01.Evaluation guide

Who it suits

Open-weight models from many labs behind one prepaid API: Kimi, Qwen, DeepSeek, GLM and Gemma, from $0.09 per million input tokens, with dedicated GPU endpoints and H100 clusters at $3.99 an hour when serverless limits are not enough.

Consider something else if…

You need a free trial, fixed rate limits or a chosen region: there is no free offer, rate limits rise with traffic and are not published per model, serverless requests pick no region, and HIPAA covers dedicated deployments only.

Editorial · Palash Bagchi · approved

Each value below links to the page it was read from, with the sentence it came from. Criteria Together AI does not publish are not listed.

Together AI pricing

4 criteria
Free tierNo
Sources (1) →
Kind of free offerNo free offer
Sources (1) →
Entry paid planNo plan fee, but the platform is fully prepaid: access requires a minimum $5 credit purchase, and API access is suspended when the balance reaches zero. Purchased credits do not currently expire.
Sources (1) →
  • Credits - Together AI Docs ↗

    “Together AI does not currently offer free trials. Access to the Together platform requires a minimum $5 credit purchase. Together AI is fully prepaid. [...] If your balance reaches zero, API access is suspended until you add credits. [...] No, prepaid balance credits in your Together.ai account do not currently have an expiration date.”

    Read 2026-09-25 · official docs · Report a correction

Pricing modelFour products priced separately: serverless inference per 1M input and output tokens for each model, with a lower cached-input rate on some; dedicated inference and GPU clusters per GPU-hour; and fine-tuning per 1M training tokens, with a minimum charge per job.
Sources (1) →
  • Pricing | Together AI ↗

    “Most teams start with serverless inference and move to dedicated endpoints at scale. [...] Price per 1M tokens Batch API price Model Input output MiniMax M3 $0.30 $0.06 (cached) $1.20 [...] All prices per gpu per hour [...] Each job is subject to the minimum charge shown for its model.”

    Read 2026-09-25 · official pricing · Report a correction

Infrastructure

1 criteria
RegionsServerless requests are routed across Together's own infrastructure with no region selection; DeepSeek models are named as hosted in North America data centers. Region choice, including EU regions, comes through dedicated endpoints or an arrangement with Together.
Sources (1) →
  • Privacy and security - Together AI Docs ↗

    “Together supports private networking and VPC-based deployments, including in EU regions. Serverless endpoints do not offer region selection; use a dedicated endpoint or contact us to discuss the right setup for your workload. [...] For example, DeepSeek models are hosted in Together's secure North America data centers.”

    Read 2026-09-25 · official docs · Report a correction

Inference

8 criteria
Input price, top model$3.00 per million input tokens for Kimi K3, the most expensive model in the serverless chat table, with cached input at $0.30. Together hosts other labs' open-weight models, so "flagship" here means the top of what it serves.
Sources (1) →
Output price, top model$15.00 per million output tokens for Kimi K3.
Sources (1) →
Input price, cheapest model$0.09 per million input tokens for Qwen3.8 Flash, with output at $0.28 - the cheapest paid chat model on the pricing page. DeepSeek V4 Flash 0731 is $0.14 in and $0.28 out. Ternary Bonsai 27B is listed at $0.00 for both input and output.
Sources (1) →
Context window1,048,576 tokens on Kimi K3 and on DeepSeek V4 Pro 0813 and V4 Flash 0731; 1,048,575 on GLM-5.3 and GLM-5.2; 1,000,000 on Qwen3.8 Flash and the Qwen3.6/3.7 Plus models; 131,072 on gpt-oss-120b and Llama 3.3 70B. The window is the model's, as listed in the docs catalogue.
Sources (1) →
  • Available models - Together AI Docs ↗

    “Kimi K3 moonshotai/Kimi-K3 1048576 [...] DeepSeek V4 Pro 0813 deepseek-ai/DeepSeek-V4-Pro-0813 1048576 [...] GLM-5.3 zai-org/GLM-5.3 1048575 [...] Qwen3.8 Flash Qwen/Qwen3.8-Flash 1000000 [...] OpenAI GPT-OSS 120B openai/gpt-oss-120b 131072”

    Read 2026-09-25 · official docs · Report a correction

Max output tokensNo separate output cap is published: max_tokens is unset by default, and the model generates until it hits a stop condition or the context limit.
Sources (1) →
  • Parameters - Together AI Docs ↗

    “The maximum number of tokens the model is allowed to generate in the response. [...] Typical default: unset (the model generates until it hits a stop condition or the context limit).”

    Read 2026-09-25 · official docs · Report a correction

Prompt cachingSupported
Sources (1) →
Batch discountBatch API advertised at up to 50% off serverless rates, but only two models are listed as discounted, at 50%: Llama 3.3 70B Instruct Turbo and Whisper Large v3. Other models run at standard rates, and batch jobs on dedicated inference get no discount.
Sources (1) →
  • Overview - Together AI Docs ↗

    “You get up to 50% off serverless rates and a separate rate limit pool [...] Selected serverless models run at 50% off batch rates: Model ID meta-llama/Llama-3.3-70B-Instruct-Turbo openai/whisper-large-v3 Models not listed run at standard rates. [...] but the discount does not apply to dedicated model inference usage.”

    Read 2026-09-25 · official docs · Report a correction

Rate limitsDynamic limits set per organization and per model, rising with sustained successful traffic; no fixed per-model figures are published. Requests over the limit get a 429; dedicated endpoints are the route to a fixed limit.
Sources (1) →
  • Rate limits - Together AI Docs ↗

    “Limits are set per organization and per model, and they adjust based on your recent usage. [...] Sustained, successful traffic raises your dynamic rate over time. [...] Dynamic rate limits adjust with usage, so there are no fixed per-model limits published.”

    Read 2026-09-25 · official docs · Report a correction

GPU

2 criteria
H100 SXM, per GPU-hour$3.99 per GPU-hour for an on-demand NVIDIA HGX H100 on GPU Clusters; preemptible is $1.99, and reserved falls to $3.19 for 91-180 days.
Sources (1) →
  • Pricing | Together AI ↗

    “NVIDIA HGX H100 Promo Promotion valid until 09/30/26 $5.49 $3.99 Contact sales [...] Preemptible Compute ON-Demand Reserved Pay as you go Pay as you go 7-30 days 31-90 days 91-180 days 181+ days NVIDIA HGX H100 $1.99 $3.99 $3.69 $3.45 $3.19 Contact us”

    Read 2026-09-25 · official pricing · Report a correction

Largest GPU offered280GB per GPU on a B300 or GB300 for dedicated inference, both priced by sales. The largest card with a listed hourly price there is the B200 180GB, at $8.99.
Sources (1) →

Compliance

7 criteria
SOC 2SOC 2 Type 2; the latest report was posted to the Together AI Trust Center on June 17, 2026.
Sources (1) →
ISO 27001ISO 27001:2022 certified, announced May 29, 2026; the certificate is listed in the Trust Center.
Sources (1) →
  • Together.ai Trust Center ↗

    “ISO 27001:2022 ISO 27001:2022 Certificate [...] Announcing ISO 27001:2022 Certification Published May 29, 2026 I am very pleased to announce that Together AI is officially ISO 27001:2022 certified”

    Read 2026-09-25 · official site · Report a correction

HIPAADedicated inference deployments can be restricted to HIPAA-attested clusters with a placement compliance policy. No BAA, and no HIPAA coverage for serverless, is documented.
Sources (1) →
GDPR / data residencyNo region choice on serverless. For GDPR-driven residency, Together offers EU-region private networking and VPC-based deployments through dedicated endpoints or by arrangement.
Sources (1) →
  • Privacy and security - Together AI Docs ↗

    “For customers with data-residency, regulatory, or compliance requirements (for example, GDPR-driven EU-region deployments), Together supports private networking and VPC-based deployments, including in EU regions. Serverless endpoints do not offer region selection”

    Read 2026-09-25 · official docs · Report a correction

Role-based access controlRole-based access control at two levels: organization roles admin and developer, project roles admin and editor, each with a documented permissions table. Together says roles are still being rolled out across its products.
Sources (1) →
  • Roles & permissions (RBAC) - Together AI Docs ↗

    “Together uses role-based access control (RBAC) at both the organization and project level. [...] Organizations have two roles: admin and developer. [...] Projects have two roles: admin and editor. [...] Roles and permissions are being progressively rolled out across Together's products and services.”

    Read 2026-09-25 · official docs · Report a correction

Private networking / VPCPrivate networking and VPC-based deployments, including in EU regions, for customers with residency or compliance needs; arranged through dedicated endpoints or with Together, with no mechanism named.
Sources (1) →
  • Privacy and security - Together AI Docs ↗

    “Together supports private networking and VPC-based deployments, including in EU regions. Serverless endpoints do not offer region selection; use a dedicated endpoint or contact us to discuss the right setup for your workload.”

    Read 2026-09-25 · official docs · Report a correction

Dedicated infrastructureDedicated model inference serves a model on reserved GPUs billed per GPU-minute, with no hard rate limits beyond the hardware and the autoscaling bounds.
Sources (1) →
  • Overview - Together AI Docs ↗

    “Dedicated model inference (DMI) lets you serve a model on reserved hardware [...] No hard rate limits: You're only limited by the capacity of your selected hardware, plus the bounds of your autoscaling configuration. [...] DMI bills per-GPU-minute”

    Read 2026-09-25 · official docs · Report a correction

When these numbers change, hear about it. Sources are re-checked monthly; a repricing goes out as a short note.

Confirm by email; unsubscribe from any issue. Your address goes to Kit and nowhere else — what we do with it.

Together AI in the news

Headlines from other publications, not checked by us · collected

Selected by a search on this company, refreshed weekly. Nobody here has verified these claims — where something material is confirmed, it is written into the page above with its source, the way every other fact on this site is.

Together AI pricing questions

How much does Together AI cost?

No plan fee, but the platform is fully prepaid: access requires a minimum $5 credit purchase, and API access is suspended when the balance reaches zero. Purchased credits do not currently expire.

Credits - Together AI Docs · read

How does Together AI charge?

Four products priced separately: serverless inference per 1M input and output tokens for each model, with a lower cached-input rate on some; dedicated inference and GPU clusters per GPU-hour; and fine-tuning per 1M training tokens, with a minimum charge per job.

Pricing | Together AI · read

Does Together AI have a free tier?

No.

Credits - Together AI Docs · read

Comments

Every page here is sourced and dated.