Home

Open Source

Curated open-source AI projects with verified GitHub data and hands-on breakdowns.

Tencent Open-Sourced Simultaneous Interpretation: 3s Lag

Tencent Hunyuan open-sourced the simultaneous interpretation model Hy-SimulMT on October 26 (30B-A3B, about 30.39B parameters per the HF weights with 128 experts and 8 activated per token): real-time Chinese-English speech translation with StreamLAAL of about 3 seconds on the RealSI long-form benchmark (the project notes this is not the wait before the first translated word), a GEMBA of roughly 85-86 per the official comparison chart (scored by GPT6.1-Sol) beating the compared open models and 4 of 5 commercial APIs, and lower latency than six CATTI-certified interpreters (4.5/4.7 seconds) across an 18-recording small-scale test. Key correction: the base checkpoint outputs translated text only — speech output and voice cloning need a separate TTS service (the official recipe uses Qwen3-TTS), and the 3-second figure excludes synthesis. License in one line: Apache-2.0 per the LICENSE file with third-party exceptions (only Qwen3-ASR-0.6B) and no extra restrictions, while GitHub shows NOASSERTION due to the custom license filename. Six builds (BF16/FP8/GGUF/Interleave), an official anchor of about 57 GiB on a single H200 for BF16, GGUF minimums unpublished; tooling includes SimulInfer/SimulTrain, llama.cpp support is PR #31198 (unmerged), and single files cap at 30 seconds.

One RTX 4090, One Interactive World Model: Hy-Nomas Reviewed

Tencent Hunyuan open-sourced Hy-Nomas, an interactive video world generation model (code on October 21, weights October 23, about 8B, Apache 2.0 with third-party notices): multi-turn control via text and camera trajectories, persistent KV cache memory across segments, 54 dual-stream MMDiT blocks plus a Qwen2.5-VL-7B text encoder; output at 704x1280@24FPS (playback rate, not generation speed), with 4/6/8-step distillation variants coexisting; the 280-case Hy-Nomas-Bench is self-reported with no third-party comparison at release. A single RTX 4090 runs it via INT8 weights, an FP8 decoder, and CPU offload (about 22.5GB minimum download) — with the trade-offs stated plainly: no LoRA on INT8, sliding-window KV sacrifices long-range memory, and the FP8 kernel targets SM89 only (3090 need not apply). Cold water included: the KV cache guarantees no stable 3D scene graph, no deterministic rigid-body physics, and zero official VRAM or latency figures.

AWS Infinity: an Agent Runtime Where Waiting Costs Nothing

AWS open-sourced Infinity, a suspend-and-resume agent runtime: the code lives in the hydro-project organization (aws/ and aws/hydro both 404), Apache-2.0, written in Rust, from the AWS x UC Berkeley Hydro collaboration. The official benchmark runs 80,000 agents for 20 rounds each — 3.2 million completions — at roughly 100KB of memory per agent on a single-threaded EPYC 9R14, extrapolating to about 75,000 agents in 8GB (an extrapolation; the Raspberry Pi 5 is only a memory analogy). This deep dive covers the serverless four-piece design, the simulated-model qualifier, the community critique list (durable I/O, race transfer, 25 stars and untested by the market), and how to get started.

A 0.9B Model Just Topped the OCR Leaderboard: How Real Is It?

GLM-OCR, open-sourced by Zhipu on February 3, 2026, has only 0.9B parameters (a 0.4B CogViT encoder plus a 0.5B GLM decoder) yet scored 94.62 on the vendor-reported OmniDocBench v1.5 board, above Gemini-3 Pro's 90.33 and GPT-4o's 75.02, with official cost at roughly one-tenth of traditional pipelines. This deep dive covers the three-way license split (weights MIT, code and PP-DocLayoutV3 Apache-2.0), the MTP design worth about 50 percent throughput, three verdicts on how real the top score is (PaddleOCR-VL-1.6 has since passed it, cross-board numbers are not comparable, and the official tagline contradicts its own table), the 37.6 percent weakness on aged scans, and four ways to run it: SDK, vLLM, SGLang, and Ollama.

M3.1 Flash Weights Are Out, GitHub Comes Up Empty

On October 23, 2026, MiniMax open-sourced M3.1 Flash weights: a native multimodal MoE with 428B total and 23B active parameters, a 1M-token context, MSA model-wide sparse attention (arXiv:2606.13392), and DSpark speculative decoding replacing EAGLE-style MTP. The weights live in two places: Hugging Face at MiniMaxAI/MiniMax-M3.1-Flash (public, not gated) and ModelScope at MiniMax/MiniMax-M3.1-Flash (page and API both verified 200 on release day); GitHub has no repo named after the model (github.com/MiniMax-AI/MiniMax-M3.1-Flash returns 404; the org's 37 repos include only the kernel library MiniMax-AI/MSA). The license, per the LICENSE file itself: MiniMax Community License - free for non-commercial use; commercial use requires visible attribution Built with MiniMax M3.1-Flash; annual revenue above USD 20 million requires prior written approval (api@minimax.io), with commercial use defined broadly to cover API resale and fine-tuned deployment. Three throughput calibers coexist: over 100 trillion tokens in the first 72 hours (official announcement via media), 200T over the 13 anonymous days (official, Oct 19), and 56.7T on OpenRouter during anonymity (via Zhidx, single platform); officially, agentic usage cost fell 70 percent versus the predecessor. Ecosystem: ten chip vendors including Huawei Ascend, T-Head, and Kunlunxin plus vLLM, SGLang, and FlagOS; Kunlunxin completed Day 0 while most others are still pushing first-day integration. The only official M3-to-M3.1 Flash note: keep the core MoE, change three things - sparse attention in the first three layers, QAT, and DSpark. Division of labor: the Oct 21 unmasking piece covers the anonymous saga and pricing; this one covers weights landing, repos, and the license.

Mistral Large 4: Model Is Live, Weights and License Are Not

Mistral announced Large 4 (nicknamed Le Chonk) in public preview on October 6, 2026, with the API open the same day: 1.05T total parameters activating 52B per token (the HF repo name Mistral-Large-4.0-1T05-A52B uses this figure; 49B excludes the embedding and output layers), a 1.6B vision encoder, training across 160+ languages, and 3,800 Grace Blackwell GPUs in Mistral's own European datacenters; the API context is listed at 1M while OpenRouter and AA test at 512K. Pricing is $1.36/$0.14/$4.18 per million tokens list, with an undated half-price column of $0.68/$0.07/$2.09 on the same docs page. But as of October 23 the open weights have not landed: the HF repo is a countdown page with zero downloadable files, the license field reads coming soon, the docs render only the word Open, and Artificial Analysis records isOpenWeights as false, classifying it as proprietary. GTC Berlin on October 21 reconfirmed the end-of-October weight release via the Mistral-NVIDIA industrial alliance. Self-reported benchmarks: DeepSWE v1.1 61.7 percent, Cybench 93 percent, and an 82 percent vulnerability reproduce-and-patch score where third-party data shows Opus 5.5 refuses 98.5 percent and GPT-6 Astra 100 percent of the same tasks; AutomationBench 59.9 despite the model card's own chart giving GLM-5.3 a higher 62.2. The third-party picture differs: AA Intelligence Index 38 (Opus 5.5 at 58), Omniscience at -5.3 meaning bare recall fabricates more than it admits, and 33rd of 45 on the vals index, roughly 9-12 months behind the frontier. Community reaction splits between 'about Kimi K3 level' performance doubts, an OpenRouter price roughly 10x DeepSeek's, and genuine EU data-residency defense. Running it locally takes an estimated 400GB+ of fast memory - today only a 512GB Mac Studio M5 Ultra theoretically qualifies.

DeepInfer Open-Sourced: 155T Tokens/Day, Engine or Factory?

On October 20, 2026 DeepSeek announced full open sourcing of DeepInfer, its production-grade inference framework - the first time the official production engine is available instead of open-source parts plus guesswork. The repository deepseek-ai/DeepInfer ships under Apache-2.0 and measured 17,902 stars, 1,836 forks, and 188 open issues on launch day, while the companion token codec deepseek-recipe is MIT-licensed; the two repositories carry different licenses. Every performance figure below is vendor-reported and has no third-party reproduction yet: a single cluster serves 155 trillion tokens per day (about 38 dollars per card per hour at off-peak pricing), first-token latency of 400 milliseconds, 230 tokens per second generation, 96 percent prefix-cache hit rate, and Prefill throughput per card rising from 60,000 to 1.5 million tokens per second when caching counts. SWA Bounded Replay pins the recompute of cache misses to the last 128 tokens, which nearly halves total Prefill compute and makes cost independent of context length; DSpark speculative decoding drafts up to 5 tokens, adds another 6 percent to live throughput, and yields 14K tokens per card per second; MoE expert replicas reshuffle every 10,000 steps so the busiest card runs only 9 percent above average load. Hardware covers both NVIDIA SM100 and Huawei Ascend 950 NPUs, and setup means a Python 3.12 virtual environment, the DEEPINFER_ACCELERATOR variable, and a source build that pulls DeepSeek-V4.1-Flash-1017 weights by default. The boundaries matter: DeepSeek states that Prefill and Decode disaggregation, data-parallel load balancing, and the KVCache management layer are still not fully open because they couple too deeply to internal infrastructure - this is the engine, not the whole factory - and no official benchmark against vLLM, SGLang, or TensorRT-LLM exists. In production, 8-card servers handle Prefill while 64 to 128-card supernodes handle Decode, with KVCache kept beyond 72 hours to avoid recomputation.

Memory-Native MemT 1.0: 288B Model That Remembers, Open-Sourced

Shanghai-based MemTensor released what the team calls the first memory-native foundation model, MemT 1.0, on 2026-10-20 with weights published: MemT-1.0-Flash (288B total / 17B active / 336M dynamic memory) and MemT-1.0-Nano (36B total / 3B active / 10M dynamic memory) - the largest open parameterized-memory models to date. License check (LICENSE text verified): the GitHub API reports NOASSERTION, but the LICENSE file is PolyForm Noncommercial 1.0.0 - weights are downloadable yet the license is non-commercial; commercial use requires separate permission from the team, with a separate LICENSE-PAPER file. Several media reports say "open source"; this article follows the LICENSE text. How memory-native differs from the three traditional kinds of memory: activation memory (context window plus KV cache, gone when the session ends), parameter memory (updated only via fine-tuning, not editable per item), and external knowledge bases (retrieval-dependent) - MemT encodes history into a persistent, updatable internal state that the model reads and writes directly at every token, running parameterized memory in parallel with attention inside selected Transformer layers so memory becomes part of the forward pass; Gated DeltaNet-2 token-wise stable updates (erase/write/forget/output gates), Mixture-of-Storage for capacity expansion (dense-routed memory experts); trained via KL distillation (teacher sees full context, student sees only memory state, only memory parameters optimized) on a 34-billion-token corpus (NetEase figure). Benchmarks (vendor-reported): first and second among parameterized-memory methods on 11 memory benchmarks at 50.6 / 47.0, versus the strongest parameterized baseline Temp-LoRA at 38.1 (+12.5/+8.9); behind MemTensor's own MemOS text-memory system - fidelity loss traded for learnability, stated as-is. SWE-bench Verified 500 tasks: 66.07 percent to 69.33 (experience injected pre-execution) / 69.73 (triggered mid-execution); 3-run all-success from 54.00 to 61.6/62.8 percent; Multilingual 60.00 to 62.89/63.56 percent. Long-history efficiency (dual GPU, single concurrency, fixed 128 tokens): 1.86x end-to-end speedup at 128K with persistent storage about 1/66; 3.99x at 256K; TTFT 104-112ms versus 445ms traditional; per-session state about 40MiB versus 5.06GiB, a 130-fold cut (1/66 against KV cache, 1/130 against the full traditional scheme - both stated). Long-conversation recall: on a 65K-token conversation, precise recall at 4K, correct "I don't know" at 16K, chain-rule reasoning at 32K, and confirmed-versus-proposed discrimination at 64K. At deployment the memory state binds to a user or task trajectory, serializes across sessions, and keeps no raw text. Direction: memory-driven recursive self-improvement (RSI); combined with MemOS 3.0 for a full plaintext-plus-parameterized chain. Repos: github.com/MemTensor/MemT-1.0 (194 stars at check) plus HF and ModelScope collections. Division note: hy-memory-2-0-opensource-resource (batch 57) is a bolt-on memory framework with memory outside the model; this one builds memory into the model itself - two routes, each now open.

AI Memory Framework, Open-Sourced: Tencent's 3B Tops Opus 5

Tencent Hunyuan open-sourced all three layers of Hy Memory 2.0 on 2026-10-16: the model layer Hy-Memory-30B-A3B (30B total parameters with about 3B active per token, a sparse MoE on HuggingFace at tencent/Hy-Memory-30B-A3B, arXiv 2610.22196), the SDK layer hy-memory (2.0.3 on PyPI at writing time) and the tool layer MemHub (CLI v1.0.3, bootstrapped via the npm memhub-cli package). License check: the GitHub API reports NOASSERTION (non-standard text), but the raw LICENSE file is Apache-2.0 with a Tencent preamble and no added restrictions - this article follows the LICENSE text. Official documents mix two spellings, "Hy-Memory-30A3B" and "Hy-Memory-30B-A3B"; we follow the HF repository name. Official benchmarks (vendor-reported): first place at 55.49 in a four-way memory-system comparison, ahead of Opus 5 (54.62) and DeepSeek-Flash (53.64), while Qwen3-30B-A3B scores only 44.37; MemoryArena 55.31 percent, leading runner-up EverOS by 10.2 points; CodeMemoryBench 64.2/35.6; SWE-ContextBench 28.3; LoCoMo 72.6 percent with memory tokens cut from 1159 to 703 (-39.3 percent). Internal business validation (vendor-reported): stock-picking multi-hop recall up from 50 to 90 percent, iWiki up 48.6 percent relative. Positioning is a memory-writer workhorse (analyst take, secondary): the model specializes in writing and updating memory and can in principle sit under existing systems like Mem0 or MemOS. MemHub tool count in two versions: the README lists 5 while the official WeChat article says 10 - go by actual CLI output. Five limitations stated as-is: vendor-only benchmarks without third-party replication, preview-stage API churn, MemHub limited to Linux/macOS, a 30B-class model that is nontrivial to self-host, and thin English docs. Repository: github.com/Tencent-Hunyuan/Hy-Memory (verified reachable). Division note: ai-agent-long-term-memory-sop (2026-08-08) covers the build-your-own mem0/Letta route; this piece covers Tencent's ready-made stack.

DeepSeek Weights Are Free, Yet Agent Task Bills Fall 60 Percent

DeepSeek updated V4.1-Flash to DeepSeek-V4.1-Flash-1017 on 2026-10-17: web, app and API switched in sync (the API name deepseek-flash is unchanged; the retired names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted and billed at Flash rates), with weights published on HuggingFace (huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017) and ModelScope (modelscope.cn/models/deepseek-ai/DeepSeek-V4.1-Flash-1017) under the same MIT license. Engine swap, same chassis: architecture and config files are identical to the September 10 build; only the weights changed. Parameter counts must be kept apart: total weights are 763B (measured from the HF checkpoint) - backbone 552B, Engram conditional memory 196B, and roughly 15B of vision encoder, MLP projector and DSpark draft model - with activated parameters at 8B prefill and 16B decode, and a 510GB checkpoint across 48 safetensors shards (FP4 expert weights, mixed precision). Vendor-reported benchmarks (no third-party board has scored the 1017 build yet): Terminal-Bench 4.0 42.3 (from 31.2, Claude Code harness), ProgramBench 29.1 (20.3), HLE no tools 42.6 (36.8), DeepSWE v1.1 77.0 (74.2, Mini-SWE-V2 harness), NL2Repo-Bench 68.2 (65.4), Automation-Bench 60.5 (54.8), Agents' Last Exam 35.4 (31.8), Apex Bench 72.1 (64.6) - eleven gains across the board; Opus 5 still leads on Terminal-Bench 4.0 (51.8) and HLE (56.3), stated in the body. Pricing unchanged: off-peak 0.15 dollars per million input tokens on cache miss, 0.003 on hit, 0.60 output; peak hours (weekdays 01:00-04:00 and 06:00-10:00 UTC) double. DeepSWE per-task cost fell from about 0.50 to about 0.20 dollars (chart reading), or roughly 3.3 to 1.3 yuan per Zhidx (secondary source). The Search/Fetch APIs also opened at 0.75 dollars per thousand calls (about 5 yuan per Zhidx), temporarily free inside Claude Code and Codex integrations. Wish-style coding: trained via self-feedback reinforcement learning, generating runnable projects from scratch with minimal prompts (official demos: game development and 3D modeling), with DeepSeek Harness 0.2.1-rc.1 or above recommended and compatibility tuning for Tencent CodeBuddy and WorkBuddy (secondary source). Timeline: V4.1-Flash debuted 9-10, a near-five-hour API outage on 10-15 was followed by full refunds of October 8-14 spending, and the 1017 build shipped 10-17. Division of labor: the Ascend stack piece covers kernel and hardware adaptation, the Cline piece covers wiring a coding agent, the Harness desktop-shell piece covers desktop environments; this one sticks to the 1017 weight update itself.