Strata 0.2.19
The app and the Terminal are 0.2.19. The engines (strata-engine, the new strata-inkling shape engine, and the updated Linux SYCL strata-hy4 package) are 0.1.65, and so is Control, the program that installs and updates them. The update feeds are sequence 131 (shared) and 127 (linux-aarch64).
What changes for you
- Native Thinking Machines Lab Inkling 2 Mini support (
strata-inkling0.1.65). Run Thinking Machines Lab's 26.4B-parameter looped Mixture-of-Experts model (3.4Bactive per token /5.4Blooped-active across 48 virtual visits, up to1Mcontext, text + image inputs, OpenAI-compatible tool calling, direct non-thinking answers) locally across Windows, Linux, macOS Apple Silicon, Intel Arc, and NVIDIA DGX Spark (linux-aarch64). - Three Strata Quantization packs (
strata-engine/Inkling-2-Mini-Strata), starting from a 4 GiB graphics card:- SQ-10 (
10.47 GB/9.75 GiB): fits 100% in VRAM on 12 GiB cards, 16 GiB Macs, and Radeon 780M / Strix Point iGPUs (and runs in hybrid RAM+VRAM mode from 4 GiB GPUs). - SQ-15 (
14.35 GB/13.36 GiB): fits 100% in VRAM on 16 GiB cards and 24 GiB Macs (96.68%top-1 agreement with the original weights,0.00889test KL divergence). - SQ-20 (
19.35 GB/18.02 GiB): bit-exact lossless nativeSF4+MX8build for 24 GiB+ GPUs and 32 GiB+ unified memory (100.00%top-1 agreement and0.00000KL divergence on held-out test prompts and WikiText-2). - Vision (
Inkling-2-Mini-Vision.strata,0.64 GBlossless BF16) and built-inW=4Multi-Branch On-GPU Markov Beam TreeDSparkspeculative decoding work out of the box in both the Desktop app andstrata launch. - Direct
safetensors,.gguf, and Dual-GPU EP support:strata-inklingalso loads officialthinkingmachines/Inkling-2-Minisafetensors(+ mtp.safetensors) and community.gguffiles directly, with--gpus 2Dual-GPU Expert Parallelism and CUDA-Graph mapped-memory doorbell NUMA offload.
- SQ-10 (
- 15 ready-made
strata-inkling0.1.65 packages inshape_engines, fetched automatically on demand when you choose Inkling 2 Mini:- Windows
x86_64:cuda-cuda13,cuda-cuda12,hip-strix(gfx1151),hip-rx9070(gfx1200/gfx1201),hip-rdna3(gfx1100–gfx1150, including Radeon 780M & Strix Point iGPUs),hip-rdna2(gfx1030–gfx1032). - Linux
x86_64:cuda-cuda13,cuda-cuda12,hip-halo(gfx1151),hip-rx9070,hip-rx7900(gfx1100–gfx1150),hip-rx6000, and unified fatsycl(Intel Arc A-series & Battlemage B-series). - macOS
arm64:metal(Metal 3 & Metal 4simdgroup_matrix+ barrier-free GEMV, signed & Apple-notarized). - Linux
aarch64:cuda-cuda13(NVIDIA Grace Blackwell / DGX Spark GB10).
- Windows
- Faster Qwen3.8-Flash-Next out of the box on NVIDIA Blackwell, RTX 40, RTX 30, and AMD RDNA3 (
strata-engine0.1.65, zero environment variables needed). - Update as usual. Your existing Qwen, Hy4-Flash, and DeepSeek models, chats, and settings stay untouched.
Engine & performance highlights
- Inkling 2 Mini (
strata-inkling0.1.65):- Sparse-MoE Bayes-Optimal
W=4On-GPU Markov Beam TreeDSpark+ Zero-Logits-D2H serving (0.00000KLD): reaches up to 250.3 tok/s solo (198.9 tok/son losslessSQ-20), 11,306 tok/s prompt prefill, and 4,089.7 tok/s total output across 64 concurrent requests on RTX 5090; 154.4–214.5 tok/s on RTX PRO 5000; 131.6 tok/s (2,081.1 tok/s@c=32) on2x RTX PRO 6000(--gpus 2EP); 108.9–114.5 tok/s on Apple M5 Max 128 GB; 99.0–106.5 tok/s on Radeon AI PRO R9700 (+66%..+78%); 47.3–48.0 tok/s on Ryzen AI Max+ Strix Halo 128 GB (+46%..+64%); 74.3 tok/s (1,145.1 tok/s@c=64) on DGX Spark GB10; and 159.2 tok/s on Intel Arc Pro B70 32 GB.
- Sparse-MoE Bayes-Optimal
- NVIDIA Blackwell (
sm_120/sm_121), Ada (sm_89), & Ampere (sm_86) (strata-engine0.1.65):- Native
sm_120adense_fp4_tc(192.6 TFLOPS),moe_fp4_tcv4, persistentINT8Tensor Coredense_i8_tc,vhead_bitplane2-stage LM head,row_dot2_80_sub16, andQ5_1use_sub8automatic defaults (0.000 ptsforce_checkparity): - RTX 5090 (
Qwen3.8-Flash-Next):IQ3_Sprefill +76% (3,205.9 ppts, 2.91x faster TTFT) and decode 161.6 tok/s (+19%);UD-Q4_K_XLdecode 115.9–117.6 tok/s (+36%..+46%);NVFP4prefill +36% (2.08xfaster TTFT) and decode 155.5 tok/s (+15%). - RTX PRO 5000 72 GB:
IQ3_Sprefill +72% (5,093–6,552 ppts, 2.96x faster TTFT) andagent8kdecode 265.3 tok/s (+28%). - RTX 4090 (
sm_89):IQ3_Sprefill +39% / decode +20%..+31% (117.1 tok/s);UD-Q4_K_XLprefill +42% / decode +47%..+59% (87.7 tok/s). - RTX 3090 & 2x RTX A4000 (
sm_86):IQ3_S/UD-Q4_K_XL/IQ2_XSprefill +12%..+26% (1.72xfaster TTFT) and decode +13%..+20% (100%bit-exact).
- Native
- AMD RDNA3 (
gfx1100–gfx1151) (strata-engine0.1.65):NVFP4Wave32 FP16 WMMA prefill (42–44xlower approximation error),native_h11p_kernel(PF_F16=2), discretemoe_p1aFL=3,IQ2_XS&NVFP4single-launch bitwise-exact LDS decode (em0=8),3.10xWave32 WMMADFlash2, WDDM coherent host staging, OCuLinkPcieAuto, and automaticgfx1151Strix Halo Windows defaults — delivering 1.44x–2.10x faster prefill and +17%..+22% faster decode (97.9 tok/s, peak103.7 tok/s) on RX 7900 XT.
- Intel Arc SYCL (
strata-hy40.1.65): fixes Battlemage (Arc Pro B70) XMX SYCL dispatch on Hy4-Flash.
Community acknowledgements
We gratefully thank Baptiste Rajaut (@baptisterajaut, PR #6520), @merbanan (PR #6520 stack), @cobra91 (PR #6506), @mercant1982, and Peasant Smith, OscarSheng, alesha-pro, and the Project Maya contributors whose MIT-licensed work on official safetensors / .gguf loading, InkDoorbell mapped-memory NUMA CPU offload, quantized KV modes, setup contracts, and --gpus 2 Dual-GPU Expert Parallelism helped shape Strata's unified strata-inkling support.
Updating
- App: the in-app update offers 0.2.19, or download the installer below.
- Terminal: run
strata update. - A clone of the repository: run
UPDATE.bat(Windows) or./setup.sh --update.
Downloads
Total: 50 files, 9,999,878,199 bytes.
Unchanged strata-hy4 0.1.64 packages stay on the 0.2.18 release, the two DFlash2 drafter packs stay on the 0.2.16 release, and the frozen DeepSeek 0.1.59 engines stay on the 0.2.13 release; the signed update feed downloads them automatically from their respective releases.
| File | Bytes |
|---|---|
| install-manifest.txt | 545 |
| install.ps1 | 16,236 |
| install.sh | 12,331 |
| latest.json | 2,134 |
| SHA256SUMS.txt | 5,350 |
| Strata-0.2.19-linux-x86_64.AppImage | 300,542,456 |
| Strata-0.2.19-linux-x86_64.AppImage.sig | 428 |
| Strata-0.2.19-macos-arm64.dmg | 161,211,245 |
| Strata-0.2.19-windows-x64-setup.exe | 109,463,499 |
| Strata-0.2.19-windows-x64-setup.exe.sig | 428 |
| strata-control-0.1.65-linux-aarch64.tar.gz | 78,458,819 |
| strata-control-0.1.65-linux-x86_64.tar.gz | 94,518,092 |
| strata-control-0.1.65-macos-arm64.tar.gz | 45,691,587 |
| strata-control-0.1.65-windows-x86_64.zip | 67,721,651 |
| strata-engine-0.1.65-linux-aarch64-cuda-cuda13.tar.gz | 582,724,655 |
| strata-engine-0.1.65-linux-x86_64-cuda-cuda12.tar.gz | 772,153,289 |
| strata-engine-0.1.65-linux-x86_64-cuda-cuda13.tar.gz | 584,031,359 |
| strata-engine-0.1.65-linux-x86_64-hip-halo.tar.gz | 265,735,598 |
| strata-engine-0.1.65-linux-x86_64-hip-rx6000.tar.gz | 190,489,895 |
| strata-engine-0.1.65-linux-x86_64-hip-rx7900.tar.gz | 271,188,189 |
| strata-engine-0.1.65-linux-x86_64-hip-rx9070.tar.gz | 633,044,600 |
| strata-engine-0.1.65-linux-x86_64-sycl.tar.gz | 188,035,148 |
| strata-engine-0.1.65-macos-arm64-metal.tar.gz | 2,703,157 |
| strata-engine-0.1.65-windows-x86_64-cuda-cuda12.tar.gz | 743,821,368 |
| strata-engine-0.1.65-windows-x86_64-cuda-cuda13.tar.gz | 490,092,153 |
| strata-engine-0.1.65-windows-x86_64-hip-rdna2.tar.gz | 166,003,861 |
| strata-engine-0.1.65-windows-x86_64-hip-rdna3.tar.gz | 218,297,369 |
| strata-engine-0.1.65-windows-x86_64-hip-rx9070.tar.gz | 535,921,808 |
| strata-feed-linux-aarch64.json | 4,481 |
| strata-feed-linux-aarch64.json.sig | 151 |
| strata-feed.json | 38,126 |
| strata-feed.json.sig | 151 |
| strata-hy4-0.1.65-linux-x86_64-sycl.tar.gz | 95,499,959 |
| strata-inkling-0.1.65-linux-aarch64-cuda-cuda13.tar.gz | 550,822,101 |
| strata-inkling-0.1.65-linux-x86_64-cuda-cuda12.tar.gz | 599,270,850 |
| strata-inkling-0.1.65-linux-x86_64-cuda-cuda13.tar.gz | 432,654,826 |
| strata-inkling-0.1.65-linux-x86_64-hip-halo.tar.gz | 93,472,505 |
| strata-inkling-0.1.65-linux-x86_64-hip-rx6000.tar.gz | 93,731,666 |
| strata-inkling-0.1.65-linux-x86_64-hip-rx7900.tar.gz | 95,697,377 |
| strata-inkling-0.1.65-linux-x86_64-hip-rx9070.tar.gz | 94,073,478 |
| strata-inkling-0.1.65-linux-x86_64-sycl.tar.gz | 90,185,285 |
| strata-inkling-0.1.65-macos-arm64-metal.tar.gz | 646,425 |
| strata-inkling-0.1.65-windows-x86_64-cuda-cuda12.tar.gz | 571,479,044 |
| strata-inkling-0.1.65-windows-x86_64-cuda-cuda13.tar.gz | 411,426,893 |
| strata-inkling-0.1.65-windows-x86_64-hip-rdna2.tar.gz | 51,469,645 |
| strata-inkling-0.1.65-windows-x86_64-hip-rdna3.tar.gz | 53,428,558 |
| strata-inkling-0.1.65-windows-x86_64-hip-rx9070.tar.gz | 51,802,199 |
| strata-inkling-0.1.65-windows-x86_64-hip-strix.tar.gz | 51,235,267 |
| Strata.app.tar.gz | 161,051,558 |
| Strata.app.tar.gz.sig | 404 |
Strata is free and open source. If it runs well on your PC, a coffee keeps the work on it going: