Search “Multimodal”

English translation is being prepared.

Ant Group's inclusionAI publishes the sglang_ling_v3 repository on GitHub

Ant Group's inclusionAI launches the sglang_ling_v3 repository on GitHub. Its contents are based on the open-source inference framework SGLang, which targets large language models, vision-language models, and diffusion models, and is optimized for agent workloads, large-scale serving, and RL rollout. The repository README shows that SGLang supports GPU, TPU, NPU, CPU, and Apple Silicon, and includes the built-in image and video generation engine SGLang Diffusion.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Fireworks Launches Free Weights Club, Free SFT Training for Muse Glimmer 30B

Fireworks AI launches the Free Weights Club: SFT training for Muse Glimmer 30B is free until November 30, with no credit card required; inference and evaluation calls are billed separately. The tuned Glimmer trained on Kalshi market data achieved 58.3% prediction accuracy on unseen topics, 11.2 percentage points higher than Jev's 47.1%; in a historical backtest, a hypothetical $1,000 grew to $1,071 after simulated fees, while both base Glimmer and Jev posted losses. In another synthetic benchmark, the System One serving stack achieved a median server-side processing speed 2-3 times faster than standard BF16 chat completions.

prime-rl 0.10.0 Released: Faster Training, New RL Loss IPO, and a Unified Evaluation Stack

Prime Intellect has released prime-rl v0.10.0 and verifiers v0.4.0, adding the RL algorithm IPO and setting it as the default, replacing the previous DPPO-based default loss. In tests comparing Qwen3-235B-A22B SFT against torchtitan, prime-rl achieved a 1.15× speedup (3732 vs 3253 tokens/s/GPU), and the MFU of full GLM-5.3 SFT rose from 22.3% to 40.1%. The evaluation CLI has moved from verifiers into prime-rl, with added support for models such as DeepSeek V4 Flash, Qwen3.8 Flash Next, and Nemotron-H.

English translation is being prepared.

MiniMax M3.1 Flash released: output 200+TPS, price halved

MiniMax released M3.1 Flash, a lightweight model positioned for Coding/Agent optimization, with a 1 million token context, output speed reaching 200+TPS in everyday scenarios, with P50 stable at 170+TPS. Its input price is 1 yuan per million tokens, output 4 yuan per million tokens, and cache read 0.1 yuan per million tokens, roughly half of DeepSeek V4.1 Flash-1017's peak-hour price. On Automation-Bench v1.0.6 it scored 62.2 points, surpassing DeepSeek's 60.5 points; on DeepSWE v1.1 it scored 73.7 points, lower than DeepSeek's 77.0 points.

The iPhone Duo that sold out in one minute exposes Apple's foldable screen production capacity shortfall

The iPhone Duo opened pre-orders at 8 p.m. on October 16, with inventory across all channels selling out within one minute, over 3 million reservations on JD.com, official website shipments scheduled as late as January 2027, and industry estimates that first-batch volume across all channels does not exceed 200,000 units. The article attributes the shortage to the 0.01-millimeter-level tolerances of the self-developed hinge, industry yields for foldable OLED panels of less than 70%, and whole-device assembly yields of just over 60% as of September 17. Counterpoint expects Apple to reach second place globally in its first year with up to 6 million units and a 25% share, while Huawei drops to third with about 5 million units and 22%, and global foldable shipments in 2026 are expected to grow only 0.5% to 20.2 million units.

TinyFish Makes Search Free: Breaking Down the Real Cost of AI Agents Reading Web Pages

TinyFish announced that, within limits, it will waive retrieval fees for Search and Fetch: Search is limited to 30 requests per minute and 500 per hour, while Fetch is limited to 150 URLs per minute and 1,000 per day. The article notes that free retrieval does not eliminate the inference costs of downstream models; the same article of about 4,300 characters returns an estimated 1,184 to 42,774 tokens across different providers, and three pages of unfiltered content can fill a 128K context window. In the September 28, 2026 Artificial Analysis snapshot, TinyFish Search scored 71 at $9.30 per 1,000 tasks, the lowest total cost among the displayed configurations but also the lowest score, with about 59 seconds per task.

English translation is being prepared.

CAICT, Lenovo, and the AIIA Data Committee Release the "Guidelines for Building High-Quality Datasets for Agent Capability Construction (2026)"

The Artificial Intelligence Research Institute of the China Academy of Information and Communications Technology (CAICT), together with Lenovo (Beijing) Co., Ltd. and the Data Committee of the Artificial Intelligence Industry Alliance (AIIA), released the "Guidelines for Building High-Quality Datasets for Agent Capability Construction (2026)," which proposes an overall framework, key methods, and implementation paths for building high-quality industry datasets oriented toward agent capability construction. Centered on six core capabilities—environmental perception, task planning, tool invocation, multi-turn interaction, security and controllability, and industry application—the guidelines analyze the data types and construction priorities required for each capability, and provide references based on practices in fields such as enterprise operations and maintenance, healthcare, communications, intellectual property, automobiles, industrial parks, and scientific research.

What Is a Native Audio Model? How Speech-to-Speech AI Skips the Transcription Step

Native audio models process or generate acoustic representations directly, preserving information such as timing, intonation, and speaker cues that text transcription would discard. They are fundamentally different from the "transcribe to text first, then synthesize speech" pipeline: the latter changes the causal chain, and the evidence of success, the sources of cost, and the approach to risk control all differ. The article breaks down its workflow into five stages: converting waveforms into audio tokens, modeling speech and non-speech context, inferring semantics and dialogue state, generating audio tokens, and decoding natural speech.

Elorian Releases Its First Model Glimpse, Using a Visual Draft Space for Image Reasoning

Elorian has released its first model, Glimpse, opening the API as a research preview, with a waitlist for the chat interface. During reasoning, the model can crop image regions, extract video frames, or run code to measure and compare what it sees, which Elorian calls a visual draft space. Glimpse is priced at $0.60 per million input tokens and $3 per million output tokens, and according to Elorian's evaluations it surpasses GPT-6 Astra on scientific chart reasoning and short video description. The company says the model was trained on several hundred GPUs, a scale notably smaller than the frontier systems it benchmarks against, and it initially targets enterprises with large amounts of video, such as robotics, research laboratories, and media and creative firms.

Google Releases Gemini 4 Argon, Limited-Time Half-Price API

Google has released its flagship model Gemini 4 Argon, focused on coding, professional knowledge work, and sustained execution of complex tasks. The API launch price is $2 per million input tokens and $10 per million output tokens, with the discount running through November 30; from December 1, prices return to $4 and $20. On the third-party Artificial Analysis Intelligence Index, Argon High ties with GPT-6 Astra Max and Claude Fable5.1 Max at 53 points, at an average cost of $1.99 per task, about 39% lower than Astra. Google says that on an ADK project it worked unattended continuously for 11 hours, proposed six optimization plans involving 32 files, and ran 16,892 tests, two of which were merged, reducing long-session memory preparation overhead by about 40%.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Reflection's Open-Weight Model Beam Launches on Baseten: 501B MoE with 1M Context

Reflection's first open-weight model, Beam, is now available on the Baseten platform, with day-0 support through Baseten Model APIs and Dedicated Inference. Beam uses a MoE architecture with 501B total parameters and 23B activated per token, a context window of up to 1 million tokens, and is aimed at coding and agent workflows. During training, Reflection used longer RL rollouts and added a length penalty to encourage the model to complete tasks with fewer tokens; users can also adjust reasoning effort via a parameter to control how much reasoning the model invests.