Tech Meridian ← ENTITY INDEX
PROMY MERIDIAN RU

COMPANY · ENTITY #2194

Apache 2.0

Related event timeline, sources and context from the news index.

EVENT TIMELINE

16

STARTUPS · 3 SOURCES · WIRED AI · TechCrunch AI · The Decoder

Recursive unveils Flint, a 40B coding model developed using its 'Eureka Machine'

Recursive, a startup led by Richard Socher and co‑founded by Jeff Clune, is offering Flint, an open‑weight 40‑billion‑parameter coding model the company says was largely developed and optimized by its recursive AI system called the Eureka Machine. Recursive shared benchmark claims that Flint outperforms peers on coding tasks and has raised $650 million at a $4.65 billion valuation from investors including Google Ventures, Greycroft, Nvidia, and AMD Ventures.

8.0

MODELS · 1 SOURCE · Hugging Face

NVIDIA releases Kumo Forecast: pretrained MOMENT-1-large time‑series forecaster with retrieval and interpretability

Kumo Forecast (formerly NV-Tesseract Forecasting) is a multivariate time‑series forecasting library from NVIDIA that pairs a pretrained MOMENT-1-large Transformer encoder (≈340M-parameter base encoder) with a trainable forecasting head and an optional Domain-Aware Representation and Retrieval (DARR) pathway that retrieves historical patterns via nearest neighbors. The Apache-2.0–licensed Kumo-TS repo provides model code (NVIDIA/Kumo-TS), weights (nvidia/Kumo-Forecast), a DataFrame-first API, configurable horizons, an interpretability framework (lag/feature attributions, trajectory diagnostics, optional embedding Integrated Gradients), and examples for fine-tuning and retrieval-enhanced evaluation.

7.0

MODELS · 1 SOURCE · InfoQ AI, ML & Data Engineering

Google releases EmbeddingGemma 2 — compact multimodal embeddings for text, code, images, video, and audio

Google has released EmbeddingGemma 2, an open-weight multimodal embedding model based on Gemma 4 and licensed under Apache 2.0. The model produces unified 768-dimensional embeddings for text, code, images, video and audio, scales from ~270M to ~740M parameters depending on loaded encoders, supports Matryoshka Representation Learning (MRL) for truncated vectors, and is positioned for low-latency on-device search and retrieval; it is available on Hugging Face.

8.0

MODELS · 1 SOURCE · AWS Machine Learning

Reflection releases Beam 1.0 — a 501B-parameter open-weight agentic coding model available on AWS

Reflection AI released Beam 1.0, an open-weight sparse mixture-of-experts model (501 billion total parameters, 23 billion active parameters, 1M token effective context) for agentic coding and long-horizon workflows. Beam 1.0 weights are published on Hugging Face under an Apache 2.0 license and the model is available via Amazon SageMaker JumpStart in US East (N. Virginia and Ohio) with deployment options including JumpStart real-time endpoints, SageMaker HyperPod Inference, and self-managed runtimes via SGLang or vLLM; checkpoints are offered in BF16, FP8, and NVFP4 formats.

8.0

MODELS · 3 SOURCES · Google · The Decoder · Google DeepMind

Google releases EmbeddingGemma 2 — open 740M multimodal on-device embedding model

EmbeddingGemma 2 is a 740M-parameter, Apache 2.0–licensed multimodal embedding model from Google that natively maps text, images, audio, video and code into a single embedding space and is optimized for on-device use. It is built on the Gemma 4 architecture, supports modular encoder configurations, an 8K-token context window, Matryoshka Representation Learning for configurable vector sizes, and is available on Hugging Face and Kaggle.

7.0

MODELS · 3 SOURCES · TechCrunch AI · The Decoder · WIRED AI

Reflection unveils Beam, a 501B-parameter open-weight model claiming frontier performance at lower compute

Reflection AI announced Beam, a text-only, open-weight mixture-of-experts model with 501 billion total parameters (23 billion active), a 1 million token context window, and pretraining on 23.8 trillion tokens. The startup says Beam matches top Chinese open models on advanced reasoning benchmarks while using '3–4x less inference compute,' will publish weights and full technical details this month, and is targeting enterprises and sovereign 'AI factory' deployments; the performance claims have not been independently verified.

8.0

MODELS · 1 SOURCE · The Decoder

Aleph Alpha releases Kolibri, a 78B German–English MoE model with 1M-token context and open weights

Aleph Alpha has released Kolibri, a 78-billion-parameter German–English mixture-of-experts model that activates about 3 billion parameters per token, supports context windows up to one million tokens, and was trained on 768 B200 GPUs in Germany and Finland. The company says German makes up 21.3% of the training data, that Chinese models were used to generate synthetic data, and that Kolibri sits on the Pareto front of quality and operating cost; the weights are published under an Apache 2.0 license on Hugging Face and the model is positioned for public administration, aviation, and industry with EU AI Act compliance in mind.

8.0

CODING · 1 SOURCE · InfoQ AI, ML & Data Engineering

Pizza Bot: AWS engineers open-source an inbox-style, local-first agent runner

AWS engineers Joseph Dolivo and Igor Fil published Pizza Bot as an Apache-2.0 open-source, self-hosted application that lets AI agents run background workflows and return results via an email-style inbox. The client-server tool supports scheduled and webhook triggers, delegation to specialist workers, multiple model providers and MCP servers, and stores agent state and data locally under user control.

6.0

CODING · 3 SOURCES · The Verge AI · TechCrunch AI · The Decoder

Meta open-sources Muse SDK so hobbyists can build Muse hardware gadgets

Meta has open sourced code and SDKs for Muse, its AI agent, allowing developers and hobbyists to run Muse on devices like Raspberry Pi, ESP32-based boards, E Ink displays, HDMI sticks or small touchscreens. Meta is also producing 5,000 Muse Home Link units with a waitlist and warns builders to "proceed at your own risk."

6.5

COMPANIES · 1 SOURCE · NVIDIA Developer

NVIDIA publishes OpenShell and Open Agent Safety Platform for continuous in-silicon agent monitoring

NVIDIA introduced OpenShell, an Apache-2.0 open-source secure runtime for sandboxed autonomous AI agents, and described the NVIDIA Open Agent Safety Platform that pairs OpenShell on Vera CPUs with NVIDIA Sentry on BlueField-4 DPUs to provide layered, out-of-band monitoring and real-time policy enforcement. The platform emphasizes verifiable policies, out-of-band enforcement, controlling the path to the model, scaling agent authority with reasoning visibility, and a shared responsibility model across labs, enterprises, and hardware providers.

8.0

MODELS · 1 SOURCE · Cohere

Cohere releases open-source "Transcribe Arabic" ASR model claiming best open-weight Arabic accuracy

Cohere published Transcribe Arabic, an open-source Arabic automatic speech recognition (ASR) model (Apache 2.0) based on its 2B ASR work. Cohere says the model achieves a 25.87 WER on the Hugging Face Arabic ASR leaderboard, outperforming Meta’s OmniASR-LLM-7B and OpenAI’s Whisper Large V3, and was preferred to Whisper in about 96% of human evaluation tests; weights are available on Hugging Face and via the Cohere API/Model Vault.

8.0

MODELS · 1 SOURCE · Cohere

Cohere releases North Mini Code — open-source 30B (3B active) agentic coding MoE model

Cohere has open-sourced North Mini Code, a mixture-of-experts agentic coding model sized at 30 billion parameters with 3 billion active, released under an Apache 2.0 license. The model — available on Hugging Face, OpenRouter, and Cohere's Model Vault in bf16, fp8 and w4a16 formats and compatible with OpenCode — is optimized for agentic developer workflows and reports a 33.4 score on the Artificial Analysis Coding Index, higher throughput (up to 2.8x) and ~30% better inter-token latency versus Devstral Small 2 in Cohere's tests.

8.0

MODELS · 1 SOURCE · Mistral AI

Shieldstral releases 3B open-weights policy-adaptive multimodal safety classifier under Apache 2.0

Shieldstral published a 3B-parameter open-weights multimodal safety classifier (Apache 2.0) that accepts plain-language policies at inference and returns a calibrated yes/no safety score; the project claims it matches or outperforms open guard models up to 7× its size on text safety and sets a new state of the art on multimodal moderation while running on a single 16GB NVIDIA GPU. The release, accompanied by a technical report, is announced as part of the Open Secure AI Alliance (including NVIDIA).

8.0

MODELS · 1 SOURCE · Mistral AI

Mistral releases Small 4: 119B MoE multimodal model with 256k context (Apache 2.0)

Mistral announced Mistral Small 4, a 119B-parameter hybrid Mixture-of-Experts model (128 experts, 4 active) that accepts text and image inputs, offers a 256k context window, and includes a configurable reasoning_effort parameter; it is released under the Apache 2.0 license. The company says Small 4 unifies capabilities from its Magistral, Pixtral, and Devstral lines, targets chat, coding/agentic, and complex-reasoning use cases, claims substantial latency and throughput gains versus Mistral Small 3, and is available across vLLM, llama.cpp, SGLang, Transformers and other runtimes.

9.0

MODELS · 1 SOURCE · Mistral AI

Leanstral: open-source Lean 4 proof-engineering agent (Leanstral-120B-A6B) released under Apache 2.0

Leanstral, an open-source code agent trained for Lean 4 proof engineering and shipped as Leanstral-120B-A6B, is released with Apache 2.0 weights, an agent mode for Mistral vibe, a free API endpoint, a forthcoming tech report, and a new evaluation suite called FLTEval. The model uses a highly sparse architecture (6B active parameters), supports MCPs via vibe (optimized for lean-lsp-mcp), and is benchmarked on realistic repository tasks where it shows cost and efficiency advantages versus several closed- and open-source competitors.

8.0

MODELS · 1 SOURCE · Mistral AI

Voxtral launches Transcribe 2: Mini Transcribe V2 and Realtime (Realtime open-weights, Apache 2.0)

Voxtral released Voxtral Transcribe 2, a family of next-generation speech-to-text models that includes Voxtral Mini Transcribe V2 for batch transcription and Voxtral Realtime for live, low-latency use. Voxtral says Realtime’s weights are open under the Apache 2.0 license on the Hugging Face Hub; the models support 13 languages and provide diarization, word-level timestamps, configurable sub-200ms latency, and claimed competitive word-error rates and pricing.

8.0