Search “Model releases”

English translation is being prepared.

Ant Ling releases finance-enhanced model Ling-3.0-flash-Fin

Ant Ling releases its first finance-enhanced model, Ling-3.0-flash-Fin, which continues the Ling-3.0-flash architecture, maintaining 124B total parameters and 5.1B activated parameters. Through continued pre-training on financial corpora, domain post-training, and tool-use optimization, it strengthens the handling of complex financial content such as annual reports and financial workbooks. The model was tested on benchmarks including FinFIRST, FinSearchComp Verified, Finance Agent, APEX-Agents, SpreadsheetBench, and τ³-Banking, with its AA Intelligence Index v4.1.1 score rising from 38 to 41. The model will offer a one-month limited-time free API on OpenRouter, and its weights will be open-sourced next week.

Fireworks Evaluates DeepSeek V4 Pro Against Fable 5 and Kimi K3

Fireworks tested DeepSeek V4 Pro 0813 on four benchmarks: SWE-Bench Verified, LiveCodeBench v6, Aider Polyglot, and Terminal-Bench 2.1. It scored 95.2% on SWE-Bench Verified, higher than Kimi K3's 92.6% and Fable 5's 85.4%, and 92.0% on LiveCodeBench v6. On cost, DeepSeek V4 Pro spent $0.309 per task solved on SWE-bench, versus $0.808 for Fable 5; on LiveCodeBench the figures were $0.040 and $0.225, respectively. Its weakness is the multilingual breadth of Aider Polyglot, where it scored only 48.9% on Java, versus 74.5% for Fable 5. After oracle routing on a combined set of 409 tasks, accuracy was 92.4%, higher than the 86.6% achieved by Fable 5 alone, at a cost of $0.279 versus $4.510 per task, with 86.3% of tasks routed to V4 Pro.

Ant Ling Open-Sources Ling-3.0-tiny-base and Ling-3.0-flash-base

Ant Ling has officially open-sourced two Base Models, Ling-3.0-tiny-base and Ling-3.0-flash-base, and simultaneously released the weights from three stages—pre-training, mid-training, and WSM merging—for a total of 6 checkpoints. Ling-3.0-tiny-base has 7.9B total parameters and 1.3B activated parameters, while Ling-3.0-flash-base has 124B total parameters and 5.1B activated parameters; both can be used for continued pre-training, supervised fine-tuning, preference optimization, reinforcement learning post-training, and distillation. The official notes state that the Base Models have not completed instruction alignment and are not recommended for direct deployment as end-user-facing chat services without post-training.

Ant Ling Open-Sources Ling-3.0-tiny: 7.9B Total Parameters, 1.3B Activated Parameters

Ant Ling has officially open-sourced Ling-3.0-tiny, a native hybrid reasoning model with 7.9B total parameters that activates only 1.3B parameters during inference, available in BF16, FP8, and INT4 versions. The company says it scores 25 on the Artificial Analysis Intelligence Index and 16 on the Agentic Index, with output speed exceeding 160 tokens/s. The model can run FP8 inference on a single NVIDIA DGX Spark with weights of about 7.85GB, and has been adapted for local execution on Apple Silicon MacBook and Mac mini.

Meta's Muse Glimmer 30B Launches on Fireworks, Targeting Persistent Agents

Fireworks has launched Meta's Muse Glimmer 30B dense model under the Apache 2.0 license, supporting 128K+ token context and native image understanding, targeting persistent agents that require multi-turn tool calling. The model scores 75.5 on MCP Atlas, 74.6 on DeepSearch QA, and 51.2 on SWE-Bench Pro, all higher than the corresponding results of Gemma 4 31B and Qwen 3.6 27B; these benchmark figures are reported by Meta. Fireworks also provides recommended settings of temperature 1.0, top_p 0.95, and top_k 64, and supports adjusting reasoning intensity in the system prompt.

Ant Ling Open-Sources Ling-3.0-flash Hybrid Reasoning Model

Ant Ling has open-sourced Ling-3.0-flash, a new-generation native hybrid reasoning model built on a MoE architecture with 124B total parameters and 5.1B activated parameters, and is simultaneously releasing a base version along with multiple precision versions including FP8, FP4, and INT4. The model supports three deployment approaches: cloud API calls, end-to-end local inference on a single NVIDIA DGX Spark using the MXFP4 or INT4 version, and high-performance deployment achieving an average output rate exceeding 1100 tokens/s under designated GPU test configurations. On the Artificial Analysis leaderboard, its output speed is 353 tokens/s, and in the AA Intelligence Index, the weighted average invocation cost per task is approximately $0.04.

Open Model Roundup: Kimi K3, Qwen 3.8, the Distillation Controversy, and the Open-Closed Gap

The Interconnects podcast uses last week's release of Kimi K3 as the starting point for a quarterly roundup of open models, discussing its gap to the closed-source frontier, the continuing role of GLM 5.2, and progress from Chinese vendors such as Qwen, DeepSeek, and MiniMax. The episode also covers the distillation controversy, the U.S. open-model ecosystem, the distinction between frontier and near-frontier models, and cybersecurity arguments against bans.

GLM-5.2 Is a Turning Point for Open Source Agents

The author of Interconnects believes GLM-5.2 is the first open-weight model that is truly usable in coding agent frameworks, and community benchmarks show it is the only open source model able to compete head-to-head with the latest models from OpenAI and Anthropic. The author tested calling the model in Claude Code via the Fireworks API; overall capability immediately felt usable, but they encountered an issue where the framework sending images to the model caused the API session to break, requiring manual clearing of the context. The author calculates that 204 days separate the release of Claude Opus 4.5 from the release of GLM-5.2, falling within the commonly cited 6-to-9-month performance gap between open source and closed source, and believes this will intensify pricing pressure on heavy token consumers, benefiting the open source model inference and fine-tuning ecosystem.

English translation is being prepared.

Neural Algorithmic Reasoning: Teaching Neural Networks to Execute Classical Algorithms

The Gradient published an article exploring neural algorithmic reasoning, that is, using deep neural networks to capture the computations of classical algorithms such as shortest paths and sorting. The article points out that MIT's "What Can Neural Networks Reason About?" demonstrates that the better the algorithmic alignment, the better the generalization; GNNs are naturally aligned with dynamic programming and outperform architectures with weaker inductive biases on DP execution benchmarks. The author also cautions that algorithmic alignment does not mean one can arbitrarily apply any highly expressive GNN, and out-of-distribution performance remains a key challenge.