Shahid🇵🇸

building backend infra for ai apps • senior tech lead • i actually ship stuff

[Shahid🇵🇸 @shahidcodes]

Now that fire pass is sunsetting, is there any similar offering with no token limit? maybe without weekly limit?

[Shahid🇵🇸 @shahidcodes]

Sad to see such offerings go away. I’ve been enjoying this, no token meter, endless generation but i can understand it’s not sustainable. Will wait for V3 @FireworksAI_HQ

[Shahid🇵🇸 @shahidcodes]

@changloria0816 We probably need claudia as well for that

[Shahid🇵🇸 @shahidcodes]

@sixprime_ @jetbrains I get what you’re saying and agree with most parts but you emphasised HIS project in earlier reply so that has to be said.

[Shahid🇵🇸 @shahidcodes]

@sixprime_ @jetbrains well it’s HIS project, he’s free to choose or reject what type of PRs he wants right?

[Shahid🇵🇸 @shahidcodes]

@TheAhmadOsman waiting for the tool calls to be fixed

[Shahid🇵🇸 @shahidcodes]

@maximelabonne Yes, this was the first model that gave me any hope of running local models on my Mac. But the tool calls aren’t working yet, something to do with the template. Waiting for it to be resolved to test its full potential.

[Shahid🇵🇸 @shahidcodes]

Let’s go! I’ve been rooting for stepfun. It’s a great model and could be greater in future. Kudos to the team

[StepFun @StepFun_ai]

⚡️ Step 3.7 Flash is here: The new frontier is agent efficiency.

#1 ClawEval-1.1 (67.1), #1 SimpleVQA Search (79.2), #2 SWE-PRO (56.3), 95.3 on V* Python. Open weights under Apache 2.0.

Built for agentic, coding, search, and multimodal workflows — balancing speed, cost, and reliable execution.

- 400 TPS. 198B sparse MoE, ~11B active. 256K context, 3 reasoning levels.
- Understands UIs, charts, docs, images — then writes code or calls tools to act on what it sees.
- Web + visual search reaches further: more sources, deeper follow-up.
- Reliable tool use — less drift, fewer broken toolcalls. 98%+ on τ²-bench across all difficulty levels.
- Works with Claude Code, KiloCode, Hermes Agent, OpenClaw, and protocols like MCP.
- Runs locally on Mac Studio M4 Max, DGX Spark, AMD AI Max+ 395.

GitHub: https://t.co/kqlZkVIRHv
HuggingFace: https://t.co/qqceCrgPiw
GGUF: https://t.co/rR6XrnymWG
ModelScope: https://t.co/wney6Tzvqy
API: https://t.co/RvHWzRG7Fu
Blog: https://t.co/BxDiajiQ5G

[Shahid🇵🇸 @shahidcodes]

@populartourist how ? i can’t even get the tool call working

[Shahid🇵🇸 @shahidcodes]

@liquidai tool calls not working in both MLX and GGUF variable via LM Studio

[Shahid🇵🇸 @shahidcodes]

@wesleimade @ScalaWilliam @liquidai no luck still the same

[Shahid🇵🇸 @shahidcodes]

@rubenssoto_ai Daily driving kimi 2.6 turbo, pretty good. Club with Opencode Go you will have all models you need

[Shahid🇵🇸 @shahidcodes]

I don't run Claude or Codex. No subscription, no tab open, no need at all. Kimi 2.6 Turbo via Firepass and Opencode Go handle everything I throw at them, vibe coding, production bugs, work tasks. More tokens than I can burn. What are you using Claude / Codex for?

[Shahid🇵🇸 @shahidcodes]

@bridgemindai Lol, running both via OpenCode Go. It has everything I need for work tasks if I describe them well. I don’t do autonomous agentic coding—I just ask it to analyze, suggest edits, & draft plans I approve/tweak, then it executes exactly what I ask. Why would I need Claude/Codex?

[Shahid🇵🇸 @shahidcodes]

Qwen3.7-max is really really good.

Prompt: /frontend-design implement a todo app which looks like car dashboard in /tmp

Result:https://t.co/vto6uudIb3

[Shahid🇵🇸 @shahidcodes]

@0xSero I’m on Kimi 2.6 for everything including UI. I haven’t felt it lacking in any other area. Did you experience it lagging behind in backend related stuff?

[Shahid🇵🇸 @shahidcodes]

@DivyanshT91162 how does it compare to muxy which is also written in swift?

[Shahid🇵🇸 @shahidcodes]

@pranavcodes_ they are too desperate i guess. I have been facing issues for 2-3 months the internet was so bad i couldn’t make UPI payments finally switched 4 numbers to Jio. Going to go for family postpaid

[Shahid🇵🇸 @shahidcodes]

Kimi 2.6 has been my daily and exclusive model for quite sometime. I don’t really miss opus or codex.

[Kirill @kirillk_web3]

> be Kimi Founder
> 32 years old
> peers are choosing corporate jobs
> you're building AI infrastructure from scratch
> China. no English press. no Western hype.
> raise $2B. hit $20B valuation. quietly.
> nobody outside China writes your name
> invent a new optimizer. 2x more efficient than the industry standard.
> build 300-agent parallel systems. 4,000 steps. 12 hours straight.
> open source. free. beats the models everyone pays $200/month for.
> one afternoon you sit down
> record 40 minutes
> give the entire playbook away for free
> the math. the architecture. the decisions.
> a week later Western developers find it
> "wait. this exists?"
> "wait. it's free?"
> "wait. it beats Claude?"
> you were already on the next version
> they were just catching up
> different game.

[Shahid🇵🇸 @shahidcodes]

@pcshipp claude code + kimi code