Command A+ from Cohere is out now :) its our best model yet and its open source apache 2.0

@cohere hell yeah! model looks great

@cohere great work

@cohere Thank you for open-sourcing it. Very exciting model.

@ayirpelle @cohere amazing. thanks for open sourcing it!

@cohere Is Cohere still partnered with Palantir? I want to support Canadian, but not global surveillance and genocide allies. Clarity here on your positioning here would be great.

@cohere We does not have any commercial relationship with them

@cohere *do not have

@cohere Thanks for the clarification, I appreciate the quick and transparent response! We lost RIM/Blackberry after it was revealed they shared a key with the RCMP, I'd love to see Cohere maintain Canadian sovereignty where possible! Keep being awesome!

@cohere Congratulations on this milestone! Looks great 👍

@cohere Sounds great. Excited to follow Cohere’s open-source journey. I’ll be testing things out tonight.

@cohere have fun :) we welcome feedback!

@cohere I like this format

@cohere cohere dropping apache 2.0 while everyone else paywalls everything

@cohere looks solid on paper, looking forward to trying it. one of the most exciting OSS releases of the year!

@cohere Very cool release. Much needed in the space.

@cohere The model says it's not open source 🤣

@cohere what a beautiful thing to do! apache 2.0 💙

@cohere sick

@cohere Woah... How did you guys crack the hallucination rate?!?? You're more than 10 points better than the next closest model 🤯

@cohere It’s a big focus of ours. It’s very important for enterprise use. The team worked real hard

@cohere Hey Nick! I love Open source because if people like you. Amazing job! Wish you all the best!

@cohere Sick

@cohere That’s pretty exciting bro! Glad to see a new LLM rising up rising up! Keep up the good work

@cohere this is exactly how a spokesperson should communicate in 2026

@cohere Awesome! 🤗 🚀

@cohere Very nice work Nick. Keep it going. What you are doing is VERY important ♥️

@cohere Nice, I'm working on (getting ready to launch soon) and would love a fully canadian coding cli instead of codex/claude code that I can add to my Agentic Development Environment Fast is good but a 1.5T Canadian Model competing against gpt 5.5 is the dream

@cohere Really cool

@cohere very, I will try out!

@cohere People like fast things

Ranked this as the top#1 fastest model, but also flagged it as somewhat more verbose than average. The verbosity would likely inflate the benchmark score. Could this be addressed with alignment on better reward for conciseness? That would significantly improve the speed, right? Curious on your thoughts.

@cohere @elliotarledge looks like here is the comparison you asked for

@cohere great work dudes!

@cohere Insane! 🤯

@cohere Cohere always amazes me... 😙

@cohere Can I use the model inside opencode?

@cohere Awesome. Where can I download two H100s? 😁

@cohere am i missing a link? did you link something? I missed the link

@cohere Oh this is very interesting!

This looks super solid - love the focus on speed. The memory reqs are at an interesting place. Jussstt out of range of consumer devices (ie it can't run on a single rtx 6000 pro). What's your goal with this model? If it's just to be the fastest gun in the west among the server provided LLMs, aren't you then competing with Cerebras' offerings of more capable models? Or do you think there's an underserved segment of people who are looking for speed + low cost?