Skip to content
DigIntoAI/Model intelligence
Model intelligence feed

Model intelligence.

Follow the changes. Understand what comes next.

Editor’s pickOpenRouter
OpenRouter/OpenRouter; MiniMax

Why AI Router ZDR Labels Are Not a Guarantee: Inside OpenRouter's MiniMax Warning and a 4-Step Hardening Guide

Following Dragos's November 19, 2026 study of two million canary-seeded prompts, OpenRouter emailed affected customers on November 20–21 warning that prompts and outputs sent to MiniMax models from July to October 2026 may have been read by a third party—whether served directly or forwarded via AtlasCloud, Novita, or StreamLake—and removed ZDR labels from affected endpoints pending investigation.

Latest stories126

Newest published
85

OpenAI Custom Voices—Prompt Design vs Recording Clone, and Where Each Works

OpenAI’s Custom voices guide covers prompt-to-voice (type: prompt) from text and recording clone from a consent recording plus a short same-person sample. Prompt voices are free to create and save but work on GPT-Live only—not Text to Speech, Realtime, or Chat Completions. Recording voices can reach those wider APIs when the org is eligible. Docs as of October 26, 2026 (Beijing time); API product, not ChatGPT App. For reference only; prompt or recording quality (including Chinese and Japanese) has not been measured on this site.

OpenAI API / Custom voices / prompt-to-voice / recording clone / GPT-LiveRead story
86

How to Choose Gemini Video Mode—Static 1 FPS or Agentic Retrieval

On the Gemini API, video input defaults to static processing: frames at a fixed 1 FPS, placed into context in one pass. Agentic is opt-in—the model navigates the timeline from the prompt, inspects transcripts selectively, and can adjust frame rate and resolution on the fly. Set "processing": "agentic" on the video part. Only some models support it. Official efficiency and quality numbers below are Google’s own claims—for reference only; mode cost or quality has not been measured on this site.

Gemini API / video understanding / static vs agentic processing / processing: agenticRead story
87

DeepSeek Tool Calls Return 400 When reasoning_content Is Dropped

In DeepSeek thinking mode, the model returns chain-of-thought in reasoning_content next to content. When the request also carries tools, every later turn must pass back the reasoning_content from all earlier turns—even turns with no tool call. If that field is missing or stripped, the official DeepSeek API returns 400. Requests without tools do not need to pass it back. For reference only—this failure has not been reproduced on this site.

DeepSeek API / thinking mode / reasoning_content / toolsRead story
88

How Cursor Usage Limits Work on The X Subscription and Legacy Plans (October 2026)

Cursor meters usage differently on The X Subscription versus a legacy Cursor plan. The X Subscription gives paid seats one shared weekly pool for Cursor models; legacy plans use two monthly pools. Multipliers and per-tier allowances are for reference only—check the live Usage or Spending UI. Per-tier digits have not been checked on this site; for reference only.

Cursor / The X Subscription / Grok 4.7 / Composer 2.5Read story
89

Claude Max and Team Monthly Platform API Credits ($100 / $200)

Claude Max and Team plans include monthly credits. Max 5x gets $100 a month and Max 20x gets $200; Team pools $20 per Standard seat and $100 per Premium seat into one balance capped at $500. Free, Pro, and Enterprise are not eligible. After seven days on an eligible plan, link one Console organization under Billing to claim. Credits cover the Claude API, Playground, Managed Agents, and Agent SDK—not interactive Claude Code or app overages. They refresh each billing cycle, do not roll over, and stay separate from Claude / Claude Code plan limits. We have not tested claiming; for reference only.

Claude Max / Claude Team / Platform API creditsRead story
90

Same Sonnet 5.5 in Claude Code, Codex and Cursor — What Actually Differed

XDA's Joe Rice-Jones ran the same Claude Sonnet 5.5 at high reasoning in Claude Code, Codex, and Cursor on one buggy Kindle / Home Assistant WebSocket proxy—about $10 and 30 minutes each. All three scored 16/16 on his hidden final tests, but self-checks, time, cost, and open-ended edits still diverged. Access paths differed too. This is the author's one-off run, for reference only; we have not re-run it.

Claude Code / Codex / Cursor / Claude Sonnet 5.5Read story
91

Ai2 OlmoEarth 1.3 Lets You Spot Land Change Without Training a Model

On October 22, 2026 (Beijing time), Ai2 released OlmoEarth v1.3: satellite embeddings, global annual 10-meter features, and a browser Embeddings Explorer. This guide separates three jobs—year-flip comparison, a small classifier in the Explorer, and full fine-tuning in Studio—and notes coverage and license limits. We have not tested the Explorer yet.

OlmoEarth v1.3 / Embeddings Explorer / OlmoEarth StudioRead story
92

Which IBM Granite Speech Model Fits Your Recording?

On about October 21, 2026 Beijing time, IBM posted Granite Speech 5.0: a new 2B speech-language model plus updates to Turbo CTC and 4.1-2b-plus. This guide maps which recording type points to which model, keeps English-to-Mandarin as a translation direction only, and notes Apache 2.0. We have not measured accuracy, speed, or hardware needs.

Granite Speech 5.0/granite-speech-5.0-2b/granite-speech-5.0-470m-turboctc/granite-speech-4.1-2b-plusRead story
93

FastH3 Omni on a Local GPU — Thresholds, Speed Caveats, License Limits

FastVideo and Nuva Lab open-sourced FastH3 Omni for reference-guided video with synced audio. This guide maps official local GPU recipes (DGX Spark, RTX 5090, FP8/4090/ComfyUI), keeps the 5.6–11.9× claim tied to eight GB200 matched tests, and spells out the MiniMax-H3 Community License exclusions. We have not measured quality or speed.

FastH3 OmniRead story
94

Phonon Local Speech with Ollama: Speak, Export WAV, and Use the Speech API

Gradium released Phonon as open-weight text-to-speech on October 20. The smallest pack, Phonon 31m, is English-only under Apache 2.0. Besides Gradium's Python and browser paths, Ollama on Apple Silicon (v0.40.9 or later) can pull phonon for terminal speak, WAV export, and /v1/audio/speech. These packages are not a Chinese or Japanese dubbing tool; published voice names and latency figures are for reference only.

Gradium PhononRead story
95

What the DeepSeek API Costs in October 2026, Model by Model

DeepSeek's pricing page lists two chat models on its API, deepseek-flash and deepseek-v4-pro, each with a peak and an off-peak rate, plus standalone Search and Fetch APIs billed per request.

DeepSeek APIRead story
96

How Much You Can Use Kimi Code on Each Plan

Moonshot AI doesn't say how many requests or tokens Kimi Code gives you on any plan. What its documentation does describe is how the allowance works.

Kimi CodeRead story