Accelerate inference, model shaping, and pre-training on a research-optimized platform.
San Francisco, CA
Joined November 2022
- Tweets3.2K
- Following277
- Followers64.7K
- Likes2.5K
the open model stack is becoming composable
model → inference → gateway/router → harness → tools
each layer can evolve independently, so you can swap in better models, providers + tools as the ecosystem moves
we broke down the full stack ↓
together.ai/blog/the-open-so…
minimax m3.1 flash beats opus 5.5 on agentic automation at ~27x lower cost
the model has a 1m context window + vision support, and is built for coding, computer use + cybersecurity
We are LIVE with @AIatMeta to dive into Muse Spark Open!
Everything you need to know about Muse Spark Open: A conversation with Meta and Together AI xitter.cf/i/broadcasts/1MJgNbkXm…
Everything you need to know about Muse Spark Open: A conversation with Meta and Together AI xitter.cf/i/broadcasts/1MJgNbkXm…
post-training gets better the faster you can iterate
multiple lora experiments can share the same gpu capacity on together post-training, so you can test more ideas without spinning up new infrastructure for every run
together.ai/post-training
minimax m3.1 flash is now available for post-training on together ai
run sft or rl with your own data, rewards + environments, while keeping the training loop in your own python code
take m3.1 flash and shape it for the agentic workload you need
together.ai/post-training
Open models in production. Inference at real scale. Cost per token that makes finance smile.
That was the energy at GTC Berlin.
Customer and partner sessions, Kaffee Inferenz, Open Haus Berlin, and a night at Untergrund.
Thanks to every builder who came through.
Together AI retweeted
Great to see Factory using Reflection’s Beam model through Together AI.
A win-win for all three teams, and a great example of the open AI ecosystem working together.
factory got early access to beam through together before launch
they used that window to test @reflection_ai's model on real coding + agentic workflows, including tool-error recovery, conflict handling + final repo verification
by day 0, @FactoryAI already knew how beam behaved on the work that matters to their customers
factory got early access to beam through together before launch
they used that window to test @reflection_ai's model on real coding + agentic workflows, including tool-error recovery, conflict handling + final repo verification
by day 0, @FactoryAI already knew how beam behaved on the work that matters to their customers
Together AI retweeted
Everything you need to know about MiniMax M3.1 Flash → minimaxflash.com.
Now live on @togethercompute!
MiniMax M3.1 Flash weights are now public. 428B parameters total, 23B activated per token.
Agentic performance: A generational leap in agentic tasks, coding, and cybersecurity — 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym.
Efficient by design: MiniMax Sparse Attention and DSpark speculative decoding improve inference efficiency.
Hugging Face:
huggingface.co/MiniMaxAI/Min…
ModelScope:
modelscope.cn/models/MiniMax…
API: platform.minimax.io
M Plan: minimax.io/m-planMiniMax
MCode for desktop: agent.minimax.io/download
minimax m3.1 flash is now live on together ai
we’re excited to be a day-0 launch partner for the model, aka space bunny
m3.1 flash is built for agentic tasks, coding, computer use + cybersecurity, with a 1m context window and vision support
try it now together.ai/models/minimax-m…
Together AI retweeted
Better AI for global work starts with better training.
Deel’s Pangaea K1 model is trained on 43,500+ pages of expert knowledge and fine-tuned on @togethercompute.
The result: faster, more accurate answers for teams everywhere. Thanks for the partnership.
.@deel’s data science team went from setup to real rl experiments on together post-training in a couple of hours
they then ran up to five experiments at once on the same gpu capacity
more experiments per gpu means a much faster iteration loop
together.ai/post-training
.@deel’s data science team went from setup to real rl experiments on together post-training in a couple of hours
they then ran up to five experiments at once on the same gpu capacity
more experiments per gpu means a much faster iteration loop
together.ai/post-training
together post-training keeps the training loop in your hands
your code defines the data, rewards, loss, optimizer + lora or full weights. gpu setup and the hard infrastructure are already handled
start from a recipe 👇
github.com/togethercomputer/…
Muse Spark Open is the most capable US open-weight model to date, and it’s live on Together AI.
Join @AIatMeta and Together for a live technical deep dive: what Meta built, and how we serve it in production.
🗓 Mon, Oct 26 · 12 PM PT / 3 PM ET
Register → luma.com/m30txf74
greptile wants to validate every piece of software on the planet
as more code gets written by ai, the first review pass needs to be independent + fast. greptile’s v5 agent spins up sub-agents on every pr, pairing open + frontier models so one model’s blind spots don’t decide what gets caught
together ai powers part of the inference behind those open models, helping reviews come back faster and at lower cost
watch how @greptile does it
read more 👇
together.ai/customers/grepti…
sf, join us this thursday for step 5 preview launch night with stepfun
we’ll dig into the model, building + serving agent workloads, and what’s worth building as models keep improving, with dan fu, hongyu zhou, mar hershenson, saoud rizwan, joey wang
rsvp: luma.com/12t5e4lc
together post-training on b200s is $9/gpu-hour for the first eight weeks, down from $11
run sft or rl, iterate on more experiments, and keep the training loop in your own code
together.ai/post-training
