sumit

building board member @sugar_labs (advisory board @gnome), past: supercomputers @StabilityAI @carperai @join_ef @GoogleOSS @coveriga

[sumit @sumxsri]

You can now connect your ChatGPT subscription plan on your Codex tasks!

[sumit @sumxsri]

@OpenAIDevs OpenAI Codex with GPT 5.3 Codex is now in Cotyper in browser is the easiest way to use it in browser without setup.

We now have:
- GPT 5.3 Codex in Codex
- Claude Sonnet 4.6 in CC
- MiniMax 2.5 in Opencode in Cotyper
& latest models as they get released.

[sumit @sumxsri]

OpenAI Codex with GPT 5.3 Codex is now in Cotyper in browser is the easiest way to use it in browser without setup.

We now have:
- 5.3 Codex in Codex
- Claude Sonnet 4.6 in CC
- MiniMax 2.5 in Opencode
in Cotyper & latest models as they get released.

[OpenAI Developers @OpenAIDevs]

GPT-5.3-Codex is now available for all developers in the Responses API.

Start building with it today.
https://t.co/8ym3r1GTaT

[sumit @sumxsri]

it's a highly quantised llama from a few years ago

give a few years and we will have frontier models at those speeds @dhh

[DHH @dhh]

Although in my quick testing maybe this is one of those "I can make it as fast as you want, if it doesn't have to work" 😂. It couldn't arrive at a functional snake game in a single HTML file. But to be fair, it had no tools or feedback. Still blew the test!

[sumit @sumxsri]

i still write code by hand sometimes

not because an ai would not be able to do it

but because if i wrote that, i would finish it faster than explaining everything to the model

[sumit @sumxsri]

@chribjel i don't think mine would get that many anyway but still i was comparing these myself and saw your post @chribjel, so sharing it here https://t.co/6gVYjxuqvZ

[sumit @sumxsri]

Benchmarks comparing all frontier closed and open source models.
MiniMax M2.5
GLM 5
Kimi K2.5
GPT 5.3 Codex
Opus 4.6

[sumit @sumxsri]

Benchmarks comparing all frontier closed and open source models.
MiniMax M2.5
GLM 5
Kimi K2.5
GPT 5.3 Codex
Opus 4.6

[MiniMax (official) @MiniMax_AI]

Introducing M2.5, an open-source frontier model designed for real-world productivity.

- SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work.

- Optimized for efficient execution, 37% faster at complex tasks.

- At $1 per hour with 100 tps, infinite scaling of long-horizon agents now economically possible

MiniMax Agent: https://t.co/aIzrFYcfUz
API: https://t.co/fHRdSV7BwZ
CodingPlan: https://t.co/FDhZBBjQrX

[sumit @sumxsri]

how creators of minimax m2.5 created the benchmarks and evals table after testing it against opus 4.6

[MiniMax (official) @MiniMax_AI]

Introducing M2.5, an open-source frontier model designed for real-world productivity.

- SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work.

- Optimized for efficient execution, 37% faster at complex tasks.

- At $1 per hour with 100 tps, infinite scaling of long-horizon agents now economically possible

MiniMax Agent: https://t.co/aIzrFYcfUz
API: https://t.co/fHRdSV7BwZ
CodingPlan: https://t.co/FDhZBBjQrX

[sumit @sumxsri]

MiniMax M2.5 is available to use on Cotyper for free! A great fast model for coding.

[MiniMax (official) @MiniMax_AI]

Introducing M2.5, an open-source frontier model designed for real-world productivity.

- SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work.

- Optimized for efficient execution, 37% faster at complex tasks.

- At $1 per hour with 100 tps, infinite scaling of long-horizon agents now economically possible

MiniMax Agent: https://t.co/aIzrFYcfUz
API: https://t.co/fHRdSV7BwZ
CodingPlan: https://t.co/FDhZBBjQrX

[sumit @sumxsri]

@Zai_org please repost this

[sumit @sumxsri]

Introducing GLM-5 in Cotyper for free for a limited time. I think it's the best open source model for coding.

GLM-5 is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active).

Enjoy!

[sumit @sumxsri]

@MiniMax_AI please repost this!

[sumit @sumxsri]

Introducing MiniMax M2.5 in Cotyper for free for a limited time. I think it's the best open source model for coding.

In my testing, it seems that it is the fastest frontier level open source model. Faster than GLM-5

As close to Opus 4.6 as it gets in open source!

Enjoy!

[sumit @sumxsri]

try for free at https://t.co/X4FeiNGNYM

[sumit @sumxsri]

try for free at https://t.co/X4FeiNGNYM

[sumit @sumxsri]

@CheetahAi no, it doesn't need webcontainers

[sumit @sumxsri]

Introducing Cotyper + OpenCode: A new way to use free coding agents. Added OpenCode agent with Kimi K2.5, best open source coding model and agent out there.

Enjoy while it is free!

[sumit @sumxsri]

@CheetahAi sure, i will check it out

[sumit @sumxsri]

@CheetahAi thanks @CheetahAi, fixed it. can you please try again?

[sumit @sumxsri]

thanks @CheetahAi, fixed it. can you please try again?

[Cheetah AI @CheetahAi]

@sumxsri authentication is not working - fix this

[sumit @sumxsri]

it's https://t.co/X4FeiNGNYM