Devye

本地大模型报幕员 | AIGC爱好者
💻4*dgx Spark + 2*R9700 + RTX 5090 + Pro2000
本地LLM搭配
智能：Deepseek v4.1 Flash 1017
速度：Qwen3.8 27B
生图：Qwen Image 2.1

[Devye @XDevye]

@coldniko @mirai_yuki_tech 等了很久了🔥

[Devye @XDevye]

全部自动识别出来？

[Geek @geekbb]

好东西，TypeScript 写的自托管 homelab 可视化工具。

家里的服务器、虚拟机和 Zigbee 设备不用再手工画拓扑，扫一遍网段、连一下 Proxmox 和 MQTT，图自己长出来，设备在不在线还实时标在上面。

https://t.co/vSzcJScygw

[Devye @XDevye]

Google开始了新一轮的挑战，下一位是谁呢？

[Artificial Analysis @ArtificialAnlys]

Google’s Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5, and second only to Claude Opus 5.5 on the Artificial Analysis Coding Agent Index

Gemini 4 Carbon scores 59 on Intelligence Index v5, driven by leading performance in complex reasoning, knowledge, and scientific tasks. It also has large gains in coding and knowledge work, particularly when compared to other Gemini models, but it is not leading in all these areas.

As with launch of Gemini 4 Argon, Gemini 4 Carbon is being rolled out to selected users and is not yet publicly available. Google is offering a 50% launch discount, with no confirmed end date. All our charts and figures use standard pricing for both Gemini 4 Carbon and Argon.

Key results for Gemini 4 Carbon with high reasoning:
➤ Gemini 4 Carbon leads the Intelligence Index v5, with the #1 score in 4 of the 9 Index evaluations: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience, AutomationBench-AA. Carbon makes substantial gains elsewhere but is not leading in coding and knowledge work

➤ Behind only Claude Opus 5.5 on coding agents: With Antigravity CLI, Gemini 4 Carbon (high) scores 59 only 3 points behind Claude Code with Claude Opus 5.5 (max, 62) on the Artificial Analysis Coding Agent Index. Carbon ranks second on both Terminal-Bench 4.0 (64%) and FrontierCode v1.1 (51%), behind Claude Opus 5.5 (max) at 56%. Google has historically underperformed in this area and so this improvement stands out

➤ Improved agentic knowledge work, but not yet leading: On AA-Briefcase, Gemini 4 Carbon (high) scores 1609 Elo, a 126-point improvement over Gemini 4 Argon (high, 1483), but still behind models including Grok 4.7 (xhigh, 1638), Muse Wonder (max, 1659) and Kimi K3.1 Pro (1678). The largest gain comes from Presentation, up 293 Elo to 1598, alongside a 120-point improvement in Analytical Quality to 1671

➤ Lower Cost per Task than Claude Opus 5.5 for comparable overall intelligence: Gemini 4 Carbon costs $7.76 per Intelligence Index task, ~23% less than Claude Opus 5.5 (max, $10.04) but ~1.2x GPT-6 Astra (max, $6.25). It uses 123k output tokens per task, 29% fewer than Claude Opus 5.5 (max, 173k). At the same $4/$20 standard pricing, its Cost per Task is in line with Gemini 4 Argon ($7.68), while scoring 9 points higher on the Intelligence Index

➤ Leading AA-Omniscience v1.1 score, with low hallucination: Gemini 4 Carbon scores 54 on our recently updated AA-Omniscience v1.1, 6 points ahead of Claude Opus 5.5 (max, 48) and 8 points ahead of Gemini 4 Argon (46). Accuracy (% correct) rises 10 points over Argon to 58%, while its hallucination rate rises from 4% to 11%. This is still the second lowest of any model, behind only Argon, compared with 50% for both Claude Opus 5.5 (max) and GPT-6 Astra (max)

Key model details:
➤ Context Window: 1M tokens
➤ Multimodality: Text, image, video, and speech input, with text output
➤ Pricing: $4/$20 per 1M input/output tokens, the same as Gemini 4 Argon’s standard pricing, with a 50% discount. Cached input tokens receive a 95% discount ($0.20 per 1M)

Congratulations to the @GoogleDeepMind team!

[Devye @XDevye]

这是什么配置，好帅

[vivienna.btc @viviennaBTC]

三个月前还在吭哧吭哧学comfyUI，装载各种skills/CLI，还在傻乎乎地充值Seedance 抽卡，如今Claude+minimaxH3➕GPU已经可以快速直出媲美SD质量的视频。

未来，大模型解决一切。

[Devye @XDevye]

老米要玩虚的？感觉方向不太对呀

[サイバー環状線 @CyberKanjousen]

米哈游首谈 AI 技术发展，点名批评 DeepSeek🤣：

[Devye @XDevye]

@redsunschan @aidavid125 4 Spark

[Devye @XDevye]

@aidavid125 哦，好像是Qwen38FN

[Devye @XDevye]

@aidavid125 5090D 137 tps? 那我Spark岂不是白买了？

[Devye @XDevye]

震撼，比我4*Spark都快？

[Forha @Forhanvv]

Running DeepSeek-V4.1-Flash on a 12GB RTX 5070 at over 100+ tok/s is absolutely absurd. 😭

The Strata engine’s layer-major chunked reading + GSQ-RCO quantization makes consumer MoE offloading feel completely native. Prefill is flying, and more RAM directly boosts decode speed.

Local compute is getting ridiculous. 💀

[Devye @XDevye]

@Forhanvv 5090速度如何呢？

[Devye @XDevye]

代价是使用寿命减半

[Leo · Building w/ AI @leo114119]

老黄给显卡锁算力被开发者破解了：

RTX 3090 的 BF16 算力被一个内核模块解开了：87.5 → 157.4 TFLOPS，+80%。

原理：在 NVIDIA 驱动加载前，抢先往显卡的速度选择器写入 0x88888888，把消费卡的指令发射速率提到专业卡水平。

不刷 BIOS、不改驱动，nvidia 更新驱动也不用重打补丁。

实测：LLM prefill +14-36%、扩散模型、FlashAttention 提速；LLM decode 和 INT8 没变（受内存带宽限制）。

整理了一下，哪些 RTX 的显卡可以解锁更高算力：

[Devye @XDevye]

隐藏算力？

[Max For AI @MaxForAI]

爆了！！！NVIDIA RTX 20、30、40 三代显卡的隐藏 AI 算力，居然都被民间大神破解了！！！🤯

老黄锁了这么多年的 Tensor Core 算力，居然被人直接解锁了，部分显卡的计算吞吐量甚至原地翻倍！

而且最离谱的是，不需要更换硬件，不用刷 BIOS，也不用魔改 NVIDIA 官方驱动。

直接从底层解除部分 Tensor Core 运算的性能限制。

目前社区已经公布了多个型号的测试结果：

🔹 RTX 4090：165.6 → 331.6 TFLOPS，直接翻倍！

🔹 RTX 3090 Ti：BF16 从 87.5 → 157.4 TFLOPS，暴涨 80%！

🔹 RTX 2080 Ti：特定 FP16 运算吞吐量接近翻倍！

🔹 RTX 4070 Ti SUPER：FP16/BF16 运算吞吐量也接近翻倍！

甚至连 Windows 11 上都已经有人成功复现了。

要知道，RTX 2080 Ti 已经是八年前的显卡了，结果到了 2026 年，居然还有人能从里面挖出这么多计算能力。

更魔幻的是，现在连开源项目和安装教程都有了。

不同系列有对应的解锁项目：

🔥 RTX 20 系列（Turing）
https://t.co/kjb2J0nzm2
🔥 RTX 30 系列（Ampere）
https://t.co/SUw5DKBUkr

🔥 RTX 40 系列（Ada）
https://t.co/1I6Xh2U3wi
🔥 Windows 11 实验方案（改装版 RTX 4080）
https://t.co/tJoM3iAxyA

还有中文安装说明：
https://t.co/nBOW04VlHD
Linux 用户按照对应仓库的教程，安装内核模块、配置启动参数，再通过冷启动触发解锁。

NVIDIA 靠着精准的产品刀法卖了这么多年显卡，结果 GitHub 上的民间大神，硬生生把老黄的刀给撅了。

谁能想到，一张用了七八年的老显卡，居然还能在 2026 年迎来这种级别的算力解锁。

[Devye @XDevye]

和之前的有什么提升吗？

[Mia @MiaAI_lab]

Run DeepSeek V4.1 Flash 1017 with the full official weights on your 4x DGX Sparks, using TensorFold, Abliterated! 🔥

- Full support for switchless TP4
- 1M context, 16M total KV cache
- 16 concurrent sessions with 1M each
- 15 GB of free mem per Spark after load
- OFFICIAL weights, no further quantization!
- Needle passed at 100% depth
- ABLIT is off by default. Gated.

Performance tested on a switchless TP4. It should be faster with a switch.

Decode speeds, non-greedy, thinking on:
- 66 tok/s on prose, single stream
- 96 tok/s on prose, 2 concurrent streams
- 173 tok/s on prose, 16 concurrent streams

Decode speeds, greedy, thinking off:
- 85 tok/s on prose, single stream
- 113 tok/s on prose, 2 concurrent streams
- 210 tok/s on prose, 16 concurrent streams

Decode speeds per stream in deep context (128k), non-greedy, thinking on:
- 58.6 tok/s on prose, single stream
- 83 tok/s on prose, 2 concurrent streams
- 132 tok/s on prose, 8 concurrent streams

Prefill: 4000+ tok/s up to 256k.

This is the best model to run if you have 4x DGX Sparks, bar none! It works so well on the Sparks that it's unbelievable.

Get it here:
https://t.co/6Og2cWfW4f

[Devye @XDevye]

@coldniko @udofuchi 那个大的，你懂的，什么放出来呀？我刚买的2*R9700不会要吃灰了吧

[Devye @XDevye]

@mirai_yuki_tech 是不是27b？已经在用了

[Devye @XDevye]

@mirai_yuki_tech @coldniko 双R9700 = 2*R9700

[Devye @XDevye]

内存要求有点高， @coldniko 能否用Strata让双R9700+96G RAM跑Qwen3.8 Flash Next也达到200tps？

[Eliovp @Vpoile1]

Turns out there was more to get out of 3-bit. 🙂

Qwen3.8 Flash Next: close to 16-bit on 3 text benchmarks, ~259 tok/s single-request decode.
8500 tok/s prefill (64K)

2× R9700 + system RAM.
New weights are up:
https://t.co/T1i3SiahQv

[Devye @XDevye]

@X 请帮我推给所有本地大模型爱好者

[Devye @XDevye]

我的新名片😍

[Devye @XDevye]

@Vpoile1 96GB RAM enough?