After Claude account suspension, a user tests Step 5 Preview for coding
With their Claude account suspended and an appeal pending, the author tests StepFun's Step 5 Preview as a coding alternative.
In short: STEP 5 PREVIEW IS GOATED @StepFun_ai’s Step 5 Preview is very impressive. Early-access feedback was strong, and after spending a few hours testing it through Opencode, I can see why. My Claude account was recently suspended, so I've been looking for a reliable coding/agent workhorse while my appeal is pending. (FUCK anthropic.) After I spent few large sessions (1~3 hours) on opencode, Step 5 Preview might be exactly what I needed. In my own coding and agent tests so far, it's been on par with or even outperformed GLM-5.3. It's still early, but my first impressions are very positive. 600B MoE • 27B active • 1M context The best part? It's available for free for a week through several coding tools. Couldn't have asked for better timing lol. Try it while it's free: 🔹 OpenCode https://opencode.ai/docs/zen/ 🔹 Cline https://cline.bot 🔹 Kilo Code https://kilo.ai/code 🔹 OpenRouter (paid, ㅠ_ㅠ) https://openrouter.ai/stepfun/step-5-preview If you're looking for a capable coding model without burning through your API budget, this one is absolutely worth trying. I'll keep putting it through real-world coding tasks and share more findings as I go. I'm also looking forward to the open-weight release. I hope it quantizes well... And seriously, where's GLM-5.5? I need it so badly. 💀