Together AI Review 2026: Careers, Pricing, Free Tier, Founder & FAQs

Jamesty
JamestyAuthor
14 min read
Together AI Review 2026: Careers, Pricing, Free Tier, Founder & FAQs

Together AI has become one of the most talked about names in the open source AI world. It sells itself as the AI native cloud, the place where developers go to run, fine tune and train open models like Llama, DeepSeek, Qwen and Kimi without owning a single GPU. The pitch is simple. Rent the compute, skip the hardware headache, pay for what you use.

Our team at NUBIA MAGAZINE spent weeks going through Together AI's pricing pages, its careers listings, its founder history, and a long trail of user reviews scattered across Glassdoor, G2, Trustpilot and developer forums. What we found is a company with genuinely strong technology sitting on top of a confusing pricing story and a trust problem that keeps showing up in review after review. Below is everything you need to know before you sign up, apply for a job there, or plug your credit card into the platform.

Together AI Company Profile

Company Name

Together AI (Together Computer, Inc.)

Founded

June 2022

Founders

Vipul Ved Prakash, Ce Zhang, Percy Liang, Chris Re

CEO

Vipul Ved Prakash

Headquarters

San Francisco, California, United States

Industry

AI cloud infrastructure, model inference, fine tuning and training

Website

together.ai

Valuation (2026)

About $8.3 billion, following an $800 million raise

Employees

Roughly 130 to 500, depending on the source and the date checked

Free Tier Status

Inconsistent. Varies from $25 credit to $5 credit to none at all, depending on when you sign up

NUBIA MAGAZINE Rating

2.2 out of 5

Who Founded Together AI? The Founder Story

Together AI was started in June 2022 by four people with very different backgrounds who all landed on the same bet: that open source foundation models would eventually matter as much as the closed ones coming out of OpenAI and Google.

The best known name on the team is Vipul Ved Prakash, who serves as co founder and CEO. Before Together, Prakash built an open source spam filter called Vipul's Razor that ended up installed on millions of mail servers, and he later founded Cloudmark around that technology. He went on to found Topsy, a social search company that Apple acquired in 2013, after which he spent several years as a senior director at Apple. That mix of spam fighting infrastructure work and a big tech acquisition gave him a reputation as someone who understands both scrappy engineering and large scale systems.

Joining him as co founders are Ce Zhang, who serves as Chief Technology Officer and previously worked as a professor at ETH Zurich focused on distributed machine learning systems, Chris Re, a Stanford professor and MacArthur Fellow who also founded Snorkel AI, and Percy Liang, a Stanford professor who directs the Center for Research on Foundation Models and co authored the widely cited HELM benchmark. Tri Dao, the creator of the FlashAttention algorithm, joined in 2023 as Chief Scientist and is often credited as the reason Together's inference speeds are so competitive.

That is a genuinely academic, research heavy founding team, and it shows in the product. Together AI leans hard into technical credibility rather than consumer polish, which is both its biggest strength and, as you will see in the user experience section below, part of where things get messy.

Together AI Careers: What Is It Actually Like To Work There?

If you are considering a job at Together AI, the honest answer is that the picture depends heavily on which review source you look at, and the company simply does not have very many reviews yet relative to its size and funding.

One aggregator puts Together AI's overall Glassdoor style rating around 4.1 out of 5 based on a larger pool of roughly 40 reviews, with culture and values scoring well above work life balance. Another snapshot, pulled directly from Glassdoor with a much smaller sample of 6 to 8 reviews, lands closer to 3.5 to 3.7 out of 5. A separate listing with just 3 to 4 reviews shows an overall score as high as 4.7, which tells you how much a handful of reviews can swing the number at a company this size.

What employees say they love

  • Working alongside genuinely strong researchers and engineers in one of the most relevant corners of tech right now
  • Real mission alignment around open source AI rather than a closed, pay walled approach
  • Fast growth and the chance to have outsized impact on a small team
  • Competitive compensation for a company at this stage

What employees say they do not love

  • Work life balance is the lowest scoring category across almost every review source we checked, often sitting between 2.0 and 3.1 out of 5
  • Senior management and leadership transparency come up repeatedly as a weak point, with one reviewer describing leadership as a closed circle
  • As a private company, equity and when it might become liquid is unclear to staff
  • A handful of applicants on forums describe a quiet hiring process, including one account of applying to more than 100 roles without landing an interview

Together AI has grown quickly on the back of large funding rounds, and its careers page regularly lists dozens of open roles spanning data center operations, research, forward deployed engineering and go to market functions. If you are the kind of person who wants to do frontier level research and does not mind a demanding pace, reviewers consistently say the work itself is interesting. If work life balance is a dealbreaker for you, go in with realistic expectations.

Together AI Pricing in 2026: What You Actually Pay

This is where the research got genuinely frustrating, and we think that frustration is itself useful information for anyone evaluating the platform.

Together AI runs on a pay as you go, prepaid credit model. Depending on the model you choose, per token pricing runs anywhere from roughly 5 cents to 9 dollars per million tokens, with most mainstream open models like Llama, DeepSeek and Qwen variants landing somewhere between 10 cents and 4 dollars per million tokens for input and output combined. GPU rentals for training or dedicated inference sit around 3.99 dollars per H100 hour on demand, dropping to roughly 3.19 dollars an hour on longer reserved terms, with dedicated inference priced closer to 5.49 dollars per H100 hour.

Fine tuning is billed per training token, starting from around 0.34 dollars per million tokens for smaller models, with minimum job charges that scale up on larger model families. Batch processing gets a 50 percent discount compared to standard serverless pricing, which is a genuinely good deal if your workload can tolerate the delay.

Volume and startup discounts

Together AI does not publish volume discount tables on its pricing page. Instead, larger commitments route through a sales conversation. The company does run a Startup Accelerator program that, according to third party roundups, grants approved early stage companies anywhere from 15,000 to 50,000 dollars in platform credit depending on the stage of the business. That program is one of the more generous ones in the space, assuming you can get accepted.

Is Together AI Free? The Confusing Truth About the Free Tier

Here is the part that trips up almost everyone researching this platform, including us. Depending on exactly when a source checked, Together AI's free tier has been described as all of the following: a 25 dollar signup credit with no card required, a 5 dollar signup credit, and, more recently, no free trial or free credit at all, with a mandatory 5 dollar minimum prepaid purchase before the API will accept a single call.

Our best read of the timeline is that Together AI has quietly tightened its free offer over the course of 2026. Earlier in the year, new accounts appear to have received meaningful free credit, somewhere between 5 and 25 dollars, enough to genuinely test a handful of models. Billing documentation checked later in the year describes a minimum 5 dollar prepaid top up just to activate API access, with older 25 dollar and 100 dollar credit offers explicitly listed as ended.

What has stayed consistent is that Together AI keeps a short list of smaller open models, along with the FLUX.1 Schnell image generation model, available at no cost or close to it, so you can poke around the platform even on a very small balance. If a free trial with no card required is the deciding factor for you, we would recommend checking the live pricing page the day you plan to sign up rather than trusting any single review, this one included, since the policy has clearly moved more than once this year.

Real User Experience: What People Are Actually Saying

This is the section that pulled our overall score down the most. Together AI's technology gets praised constantly. The business practices around billing, support and account deprecation get criticized just as often.

The good

  • Developers on G2 repeatedly praise inference speed, calling it some of the fastest time to first token they have used for voice and latency sensitive applications
  • An OpenAI compatible API makes switching from other providers relatively painless for teams that already have tooling built around that standard
  • A genuinely large catalog, well over 200 open source models, lets teams compare quality and cost without juggling multiple vendor accounts
  • Several reviewers describe the platform as honest about pricing and responsive to feedback when things go well

The bad

  • Trustpilot shows a weak score, around 2.5 out of 5 from a small set of reviews, including accounts describing the service as a scam or using the phrase bait and switch
  • A recurring complaint on G2 and Amazon's marketplace involves serverless model instances being deprecated with little or no warning, followed by slow or refused refunds on prepaid balances
  • Documentation is frequently called thin or confusing, especially for developers without strong API experience
  • Support response times are described as slow outside of enterprise level plans
  • One independent benchmarking site scored Together AI at 2.3 out of 5 on overall quality while still rating its uptime and reliability closer to 4.0 out of 5, which lines up with our own conclusion that the infrastructure works, but the surrounding customer experience often does not

Reading through dozens of reviews, a pattern emerges. People who already know how to work with APIs, read documentation carefully and keep an eye on their account tend to have a smooth, even great, experience. People who expect a polished, forgiving consumer product, the kind where a deprecated endpoint or a billing hiccup gets fixed with one friendly support chat, tend to walk away frustrated and occasionally very vocal about it.

Nubia Magazine Verdict

We want to be clear about something before giving the final number. Together AI is not a bad piece of technology. Its inference speed is genuinely competitive, its model catalog is enormous, and it was built by a founding team with real research credibility. If this were purely a technical review, the score would sit much higher.

But a review has to account for the full experience of using a company, not just the engine under the hood. Across the sources we checked, three problems kept showing up together: a free tier policy that has shifted enough times to feel unreliable, a pattern of billing and account deprecation complaints that go unresolved for weeks, and trust scores on independent review sites like Trustpilot that sit well below what the underlying technology deserves. Add in a careers picture where work life balance and leadership transparency are consistently the weakest categories, and the overall picture for NUBIA MAGAZINE lands at a 2.2 out of 5.

Our advice is simple. If you are a confident developer who reads the fine print, keeps an eye on your balance, and wants fast, affordable access to open source models, Together AI can absolutely work for you. If you are looking for a forgiving, beginner friendly platform with a dependable free tier and predictable support, go in with caution, or at least a small prepaid balance you are comfortable losing while you test the waters.

Together AI FAQs: What People Are Searching in 2026

1. Is Together AI legit?

Yes, Together AI is a real, well funded company, valued at roughly 8.3 billion dollars after an 800 million dollar raise in 2026, with a functioning product used by a large developer base. Legit does not mean flawless, though, and our research turned up repeated complaints about billing and account deprecation that anyone signing up should know about.

2. Does Together AI still have a free tier in 2026?

It depends on when you check. Through parts of 2026, new accounts reportedly received free credit ranging from 5 to 25 dollars with no card required. More recent billing documentation describes a mandatory 5 dollar minimum prepaid purchase with no free trial at all. A small set of models and the FLUX.1 Schnell image generator have stayed free or close to free throughout. Check the live pricing page before you rely on any free credit figure.

3. Who is the founder and CEO of Together AI?

Together AI was founded in June 2022 by Vipul Ved Prakash, who serves as CEO, alongside Ce Zhang, Chris Re and Percy Liang. Prakash previously founded Cloudmark and Topsy, the latter acquired by Apple in 2013, and spent several years at Apple before starting Together.

4. Is Together AI a good company to work for?

Reviews are mixed and the sample size is still fairly small for a company of this profile. Culture, mission alignment and career growth tend to score well, often in the 4 out of 5 range, while work life balance and senior management transparency are consistently the weakest categories, frequently scoring between 2 and 3 out of 5.

5. How much does Together AI cost per million tokens?

Pricing spans roughly 5 cents to 9 dollars per million tokens depending on the model, with most popular open models like Llama, Qwen and DeepSeek variants falling between about 10 cents and 4 dollars per million tokens combined for input and output. Dedicated GPU rentals and fine tuning are billed separately.

6. Can I use Together AI without a credit card?

At various points in 2026, Together AI has allowed signup without a card attached to claim a small free credit. More recent billing policy appears to require a card or bank payment to fund the minimum prepaid balance before API access activates. This is another detail worth confirming directly on the pricing page at the time you sign up.

7. Is Together AI better than OpenAI or Anthropic?

They are not really direct competitors in the way the question implies. Together AI focuses on hosting and serving open source models at competitive prices and fast speeds, while OpenAI and Anthropic primarily sell access to their own closed, proprietary models. If your priority is low cost access to open weight models, Together AI is a strong option. If you want a single, tightly integrated closed model experience, the other providers are built for that instead.

8. Does Together AI have hidden fees or billing problems?

Multiple independent reviews describe serverless model instances being deprecated with little warning, followed by slow refunds on leftover prepaid balances. There is no evidence of deliberately hidden recurring fees, since the platform is pay as you go rather than subscription based, but the deprecation and refund pattern shows up often enough across review sites that we treat it as a real risk rather than a one off complaint.

9. What is Together AI's Trustpilot and Glassdoor rating in 2026?

Trustpilot shows a weak score of around 2.5 out of 5 from a small number of reviews, several of which describe billing disputes. Glassdoor style employee ratings vary widely by source, from roughly 3.5 up to 4.1 or higher out of 5, mostly because the total number of submitted reviews is still small enough that a few entries can swing the average significantly.

10. Is Together AI worth it in 2026?

For experienced developers who want fast, affordable access to open source models and are comfortable monitoring their own account and balance, yes. For anyone hoping for a dependable free tier, polished documentation and fast customer support, our research suggests you should proceed carefully and keep expectations modest, which is a core reason NUBIA MAGAZINE rated the overall experience 2.2 out of 5.


Share

0 Comments

Join the discussion and share your thoughts

Join the Discussion

Share your voice

0 / 2000

* Your email is kept private and never published.

No Comments Yet

Be the first to share your thoughts on this article!