Anime fan

W

[Anime fan @badboy999654]

@ggg78g89 Qwen 3.8 did better. It even added livestock and npcs with animations

[Anime fan @badboy999654]

@teortaxesTex Less than 24 hours until ds 4 preview depreciation, still no email

[OmedTheVibeCoder @OmedVibeCodes]

This completely changes how I look at AI benchmarks.

Qwen 3.8 Max Preview felt nearly unusable in OpenCode and Qwen Code.
With the Claude Code harness, it suddenly came surprisingly close to Kimi K3 and Fable 5.

Same model. Same task. Different harness. Wildly different result.

How is this even possible?

[Anshu @anshuc]

Wow, if you connect Fable to image-gen and image-to-3D APIs, it's *insane* for game dev.

It made an infinite explorable universe, with characters and lore. All assets and sound completely from scratch.

3 prompts in 2 hours, max effort. Workflow below.

[Anime fan @badboy999654]

@xuanmingzhangai Don't listen to people asking for small models. Look at Anthropic revenue, people want large token efficient models an are willing to pay for it

[Anime fan @badboy999654]

@ChujieZheng I like this model. Here's Christmas celebration scene I told it to build. Probably not in the training data, so not benchmaxxed. https://t.co/su3nzeDkPP

[Anime fan @badboy999654]

@shuai_bai_ Amazing model. I built a Christmas celebration scene in there js. https://t.co/su3nzeDkPP

[Anime fan @badboy999654]

@teortaxesTex And people still have the audacity to doubt qwen lmao

[Anime fan @badboy999654]

@teortaxesTex You should try new V4 flash on the web instead of shit posting. It's actually good. I'm having fun testing it because it's fast and actually can improve mid results when you tell it to.

[Anime fan @badboy999654]

@teortaxesTex Huawei actually has the supply chain to connect those chips together

[Anime fan @badboy999654]

@teortaxesTex Ds flash on the web is very good too. I think it has been updated.

[Anime fan @badboy999654]

@HaoningTimothy Is there a self improvement flywheel like Anthropic that enables model releases every month?

[Anime fan @badboy999654]

@teortaxesTex Stop with this bullshit. You're becoming like that Anthropic fanboy Lisan

[Anime fan @badboy999654]

@ZixuanLi_ It's coming!?

[Anime fan @badboy999654]

@exploraX_ @teortaxesTex

[Anime fan @badboy999654]

@shuai_bai_ I hope it beats Fable in scientific reasoning and visual understanding

[Anime fan @badboy999654]

@Alibaba_Qwen Please don't nerf the web version. People are saying model is bad by only testing the nerfed web version.

[Luckey Faraday @luckeyfaraday]

If you’ve tried Qwen3.8 Max Preview on OpenCode note that the effort level is the lowest by default

You have to configure this beforehand in OpenCode’s settings (just ask claude or codex to do it)

The efforts levels are none, minimal, low, medium, high and xhigh

xhigh is significantly better and yields amazing results

[Anime fan @badboy999654]

@quimedesu Are you using API? What harness?

[Anime fan @badboy999654]

@teortaxesTex V4 flash already deployed on the web. Turn off search and turn on deep think. It's not bad so far