ElevenLabs' Music v2.5 ships with lossless downloads on the free plan

ElevenLabs released Music v2.5, calling it the best version of its music generation model yet — and paired the model upgrade with a rights change that matters as much as the sound. Every track made in ElevenMusic can now be downloaded losslessly and kept, with five downloads a day on the Free plan and 400 per month on Pro, and the company says the rights to what you make attach the moment you create a track, surviving cancellation or a plan downgrade.

ElevenLabs shipped Music v2.5 today, and the quieter headline is ownership. The model itself is an iterative step: in the company's blind test — 47,885 prompt-matched pairs with two takes each, one from v2 and one from v2.5 — listeners preferred the new take the majority of the time, with the widest gains in vocal-led and acoustic-heavy genres like R&B, soul, hip hop, rock, metal, orchestral and cinematic. Music v2.5 is now the default in ElevenMusic, live in ElevenCreative and the API, and v2 isn't going anywhere for people attached to the old sound.

The commercial terms are the sharper move in a market still fighting over training data. Tracks are downloadable as lossless files, commercial use on the Free plan requires crediting ElevenMusic, and the company commits that future terms changes apply only to tracks made after the change — the songs you already made stay yours. There is one hard limit: downloads are blocked on tracks built on another artist's song, the same rule that protects the artists, labels and publishers ElevenMusic was built with — the deal structure we covered when Universal Music and ElevenLabs sign first major-label AI music deal landed last month.

The take: after a year of AI music companies treating output rights as fine print, a market leader putting "yours, commercially, on the free plan" into the default experience is a competitive statement — and a shot at every rival still hiding downloads behind paid tiers. It also quietly answers the licensing question: permission is bundled with the model, not litigated after.

What to watch: whether Suno and other rivals match the free-tier download rights, and whether the artist-reference download block holds up as the catalog of derivative prompts grows.

If AI music platforms gave you full commercial rights tomorrow, would you actually ship a track? Tell us in the comments.

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base. The run itself is the number worth keeping: 16,000 Nvidia GB200 GPUs for 38 days across August and September, at 21% average model

Read more

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute