Pinned Tweet
Introducing LingSpark: Ling-3.1-flash (560B MoE, 25B active) on 2x DGX Spark with TensorFold.
Speed: 33-35 tok/s single stream, full 1M context, EXL3 2.71 bpw (197 GB).
Quality: GSM8K 97.6% (500 q), MMLU-Pro 76.5% (200 q).
Of course it runs on TensorFold @tensorfold_dev ely no multimodal with this model. It appears great for agentic workloads and coding. It was fun porting it anyway.
Link to recipe in comments.
Speed: 33-35 tok/s single stream, full 1M context, EXL3 2.71 bpw (197 GB).
Quality: GSM8K 97.6% (500 q), MMLU-Pro 76.5% (200 q).
Of course it runs on TensorFold @tensorfold_dev ely no multimodal with this model. It appears great for agentic workloads and coding. It was fun porting it anyway.
Link to recipe in comments.
3
0
6
237
Amongst all the Qwen 4 “waiting for”, this is interesting news. A 1M flash model with vision. I will get this ported asap to Spark.
ERNIE 5.5 Flash Preview just landed! 🔥
Built on a new architecture designed for efficient inference, the latest open-weight foundation model in the ERNIE family delivers stronger performance in coding, 3D modeling, knowledge work and content creation.
It keeps decoding fast ve scores on Terminal-Bench, AutomationBench, and other benchmarks.
Here's the rundown:
- 280B total parameters, 14B active per token
- 1M-token context window
- Generation speed above 200 tokens/sec
Model weights will be released by November 25.
2
0
3
177
TensorFold 1.2.6 is out and it has my code in it. Very soon we will see a few new model families in there as well.
Thanks @ashxhart for the opportunity.
Thanks @ashxhart for the opportunity.
TensorFold 1.2.6 is out. 🚀
The big one here is the prompt cache under multi-agent load.
In the 16-agent Qwen3.8-27B + DFlash2 test, median follow-up time to first token dropped:
• M3 Ultra: 546s → 15.9s
• GB10: 105s → 25.8s
Each agent started with a ~17k-token prompt. that test.
Also in this release:
• Flash Next cold prompts finish 29 to 41% sooner on HIP, measured on the Radeon 8060S.
• GLM-5.3-Flash single-stream code decoding is 32% faster on two GB10s at TP=2.
• Opt-in DSpine-b16 drafting from Tencent for Qwen3.8-27B on CUDA.
• An image-generation API for image-capable engines, local Unix sockets and tool-call fixes.
One upgrade note: kept prompt states need a fresh prefill after updating.
Massive thanks to everyone contributing code, testing on their hardware and reporting the things that need fixing. I really appreciate the time you all put into this.
@andersjw_ @Bizuayeu @defilan @devopsfortunato @unigilby @jvr0x @teleogen
And on GitHub: abm9111, AbrahamCete, akol1, BobClawblaw, chadhurley25075-png, davidtai, ElderOrb, jschmied, kotevcode, Mirrdhyn, mrv777, ppardus, robertung88 and volfco.
@petruspennanen @volatilemarkts @plotarmordev @MiaAI_lab
tensorfold.dev
3
1
11
529
This machine … is awesome.
A datacenter in every home.
A customer emailed us this photo today. Their new Autonomous Computer, built with 2x RTX PRO 6000 Blackwell GPUs, right on their desk.
Just super happy to see our builds out in the world helping people own their compute!
autonomous.ai/computer
2
0
5
251
NVMe as KV tier allows execution of a relatively large model on a modest MacBook. Very interesting inference approach.
Half a million tokens of context. 27B model. 24 GB MacBook.
I want to be prepared for the Qwen4 release.
I built a custom inference engine for Ternary Bonsai 2 27B (a 2-bit build of Qwen3.8-27B):
- 500k ctx at 19 tok/s, 15 GB peak RAM
- 2.3x faster than llama.cpp at 4x the paged KV: each decode step attends only to the top pages, picked from per-page min/max key summaries
> NVMe as a KV tier: 32 GB of cache with only 2 GB in RAM
> Custom Metal flash-decoding kernel that reads through the page table
> MTP speculative decoding: verifying 2 tokens costs about the same as decoding 1
> YaRN with mscale = 1.0; the stock value breaks 2-bit models
0
0
1
40
Muse Ember is quite an interesting model (dev.meta.ai/resources/blog…) optimised for high-volume agentic workloads. I previously ported their Spark model to Spark (yes), 816B MoE. This round should be easier and hopefully more performant. Will keep you posted as the weights are released.
1
0
1
67
New models from Anthropic, Fable 5.2 and Mythos 5.2. I had personally hoped for an Opus 5.6 as well.
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
1
0
0
62
The next model I work on needs to have vision. These agent-focused models like Ripple and Ling are great, but there is nothing better like ingesting a receipt or producing visual output.
0
0
2
100
RippleSpark (@RicursiveAI Ripple 65B) pushing code through the 2 x Spark pipe. GPUs go brrr.
1
0
2
187
LingSpark is being tested. Very proud of this recipe. Thank you, @plotarmordev
There is a 560B model running with its full 1M context on 2x DGX Sparks by @andersjw_.
His LingSpark recipe runs Ling-3.1-flash on @ashxhart's TensorFold. He measured 32 to 35 tok/s, and still about 26 tok/s after a 128K prompt.
That prompt takes about 4.5 minutes to read the urns reuse the cached conversation and start in 7 to 8 s.
Really nice work!
0
0
3
118
RIPPLESPARK: @RicursiveAI Ripple 65B on 2x DGX Spark: from 9.9 to 46.6 tok/s coding decode with DFlash2 and TensorFold, 5.0bpw. Try it out now!
Link to port in comments as well as more stats.
Link to port in comments as well as more stats.
6
0
4
357
@niklaslenz_ai Maybe Ripple should feature in one of your famous battle videos?
0
0
0
15
By context window: you really see the constraints of Sparks and model density as context window increases
0
0
0
8
First-pass port of Ripple to 2x DGX Spark for TensorFold completed, upload of model underway. Ripple 65B is a dense model with 80 layers and therefore quite slow on Spark despite its relatively small size. Don't expect a speed devil. I will share the details here as soon as the model is uploaded.
1
0
4
188