# MAI Code 2, SenseNova-Loop-9B, sparse attention, Gemini tasks

> Microsoft ships MAI Code 2 models, SenseNova opens a 9B looped model, NVIDIA maps sparse attention costs, and Gemini gets a Task mode.

Published 17 Nov 2026. Coverage: 16 Nov 2026 to 17 Nov 2026. By Ali Asif.

## Items

1. **Introducing MAI Code 2 and Code 2 Flash** (Microsoft AI, Agents and coding): Two coding models in GitHub Copilot; Microsoft reports MAI Code 2 at 98% of GPT 6.1 Sol's average benchmark score at 40% of the cost, with costs shown at a 50% promotional price. [Primary source](https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash)
2. **sensenova/SenseNova-Loop-9B** (SenseNova, Open weights): An Apache-2.0 image-text-to-text model with about 9B parameters, a looped middle block and a 512K-token context, reporting 59.55 on Terminal-Bench 2.1. [Primary source](https://huggingface.co/sensenova/SenseNova-Loop-9B)
3. **Co-Designing Sparse Attention for Efficient Long-Context Inference** (NVIDIA, Science): Sweeps the indexer knobs in sparse attention and ends with eight design guidelines, including a Top-k method 2.1x to 2.4x faster than radix select on DeepSeek V3.2 and V4 decode. [Primary source](https://developer.nvidia.com/blog/co-designing-sparse-attention-for-efficient-long-context-inference)
4. **Put Gemini to work for you** (Google, Agents and coding): Tasks in Gemini, a Task toggle in the app running on Gemini 4 Hydrogen, is free in six countries and can pay through Google Pay and AP2 with a spending cap. [Primary source](https://blog.google/innovation-and-ai/products/gemini-app/tasks-in-gemini-free-new-features)

Four items today, all with something concrete to read. Every number below comes from the pages I opened. I have not run any of it. OpenAI also posted about GPT-6.2 Astra and Sol, but its page would not load for me, so I am leaving it out until I can read it.

## MAI Code 2 and Code 2 Flash

Microsoft announced two coding models at Ignite, and both are available in GitHub Copilot. MAI Code 2 is aimed at enterprise work such as deep debugging, multi-file implementation and repo-wide reasoning. Code 2 Flash is for high-volume jobs like explanations, targeted edits and routine fixes. Microsoft says plainly that neither is tuned for very long-horizon agentic work.

The headline claim is an average over DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1: MAI Code 2 reaches 98% of GPT 6.1 Sol's score at 40% of its cost, and 96% of Claude Opus 5.5's score at 5% of its cost. In a blind frontend evaluation with 258 annotators, reviewers preferred its output over GPT-6 Sol 64% of the time. The cost figures are shown at a 50% promotional price, and all models ran at max reasoning effort. Flash is in limited preview in Microsoft Foundry.

## SenseNova-Loop-9B

This open model has about 9B parameters and an inference-equivalent size of about 14.6B. It has 32 decoder layers but runs 56 layer executions per forward pass, because layers 4 to 27 form a block that loops twice. It takes text and images, outputs text, and supports a 512K-token context under Apache 2.0.

The card reports 59.55 on Terminal-Bench 2.1, 69.90 on GAIA and 65.31 on AIDABench. On NL2Repo it scores 41.36 against 44.60 for Qwen3.8-27B. With 4-bit weights of about 4.9 GB, a 256K context needs about 15 GiB with an 8-bit KV cache. The repo is gated, so you accept access terms before downloading.

## NVIDIA on sparse attention design

At 128K tokens, NVIDIA measures dense attention at 91% of a layer's prefill time. In DeepSeek V3.2 the indexer and Top-k take 47%, core attention 24% and the feed-forward block 25%. The post sweeps indexer settings and finds, for example, that raising key heads from 1 to 8 makes decode 7.6x slower, while doubling head dimension costs only 1.1x to 1.2x in prefill. Its Guess-Verify-Refine Top-k in TensorRT LLM is 2.1x to 2.4x faster than radix select on DeepSeek V3.2 and V4 decode.

## Tasks in Gemini

Google added a Task toggle to the Gemini app, running on Gemini 4 Hydrogen. It handles multi-step jobs such as booking, calling businesses and shopping, connects to more than 60 third-party apps, and can pay through Google Pay and AP2 with a hard spending cap. It is free in the U.S., Japan, Korea, Canada, Australia and India. Google notes tasks can fail, and some sites hand the final click back to you.

## What I'd watch next

Independent runs of MAI Code 2 outside Microsoft's own benchmark set, and whether the Loop-9B numbers hold once people can load the gated weights.

## My take

MAI Code 2 is the one to test, but the cost comparison rests on a promotional price and Microsoft's own benchmark average, so I would rerun it on my own repos. SenseNova-Loop-9B is interesting because it reuses layers instead of adding parameters, though the card compares against a Qwen3.8-27B model and a preview model rather than a wide field. The NVIDIA post is the most useful reading if you build inference stacks.

## Sources

- https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash
- https://huggingface.co/sensenova/SenseNova-Loop-9B
- https://developer.nvidia.com/blog/co-designing-sparse-attention-for-efficient-long-context-inference
- https://blog.google/innovation-and-ai/products/gemini-app/tasks-in-gemini-free-new-features

Canonical: https://heyaliasif.com/ai/2026-11-17-mai-code-2-sensenova-loop-9b-sparse-attention-gemini-tasks