New on Ropewalk This Week: Veo 3.1 Lite, Lyria 3 Pro, Grok 4.5 & Kimi K2.6 (July 2026)

Four new AI models landed on Ropewalk this week: a video generator priced at $0.05 per second, a music model that writes full 3-minute songs for $0.08 a track, and two flagship text models — one from xAI, one the platform's first from Moonshot AI. Here's what each one does, what it costs in gems, and how to try it right now.

By Ropewalk Team. Tested on 2026-07-28 with live generations across all four models — a real 8-second video from Veo 3.1 Lite, a real full-length song from Lyria 3 Pro, and real text replies from Grok 4.5 and Kimi K2.6.

What shipped this week — 4 models, July 2026:

• Google Veo 3.1 Lite — Google's cost-efficient video model with native audio, $0.05/second at 720p
• Google Lyria 3 Pro — full 3-minute song generation, up from the 30-second clips of the base Lyria 3
• Grok 4.5 — xAI's current flagship text model, trained with Cursor
• Kimi K2.6 — Moonshot AI's 1-trillion-parameter flagship, the first Kimi-family model on Ropewalk

Google Veo 3.1 Lite

Google Veo 3.1 Lite is a cost-efficient version of Veo 3.1 built for high-volume video generation, and it's the first Lite tier of Veo on Ropewalk — the catalog previously only had the full and Fast variants. It renders native audio alongside the video in the same generation, runs at $0.05 per second at 720p resolution or $0.08 per second at 1080p, and supports both 16:9 landscape and 9:16 portrait aspect ratios in clips of 4, 6, or 8 seconds. We generated an 8-second 720p clip of a hummingbird hovering beside a hibiscus flower as a verification test: it completed in 37.7 seconds and billed 1,600 gems, matching the $0.05/second rate exactly.

Google Lyria 3 Pro

Google Lyria 3 Pro is Google's most capable music model on Ropewalk, and it solves the biggest limitation of the base Lyria 3 already on the platform: it composes full-length songs up to 3 minutes long instead of 30-second clips. Generation is a flat $0.08 per song regardless of length, and it also accepts up to 10 input images to steer the composition's mood. We generated an upbeat acoustic folk track — fingerpicked guitar, brushed drums, a whistled melody — as a verification test: it completed in 37.1 seconds and billed 320 gems, matching the $0.08 flat rate.

Grok 4.5

Grok 4.5 is xAI's current flagship text model, trained with Cursor and live via the api.x.ai endpoint since 2026-07-08. It carries a 500,000-token context window — the widest of the two text models added this week — and is priced at $2 per million input tokens and $6 per million output tokens. We tested it with a short verification prompt: it returned a real reply in a single pass at roughly 5 gems for that exchange (474 input tokens, 12 output tokens), confirming the model resolves and bills correctly end to end.

Kimi K2.6

Kimi K2.6 is Moonshot AI's flagship model and the first entry from the Kimi family on Ropewalk — until this week the catalog had zero coverage of Moonshot's lineup. It's a 1-trillion-parameter mixture-of-experts model with 32 billion active parameters and a 256,000-token context window, served via togetherai at $0.95 per million input tokens and $4 per million output tokens. Our verification test returned real content (474 input tokens, 348 output tokens, roughly 10 gems for that reply) without hitting the streaming-required errors that still block some other togetherai models on the platform.

New Models at a Glance

Model	Type	Price	Context / Length	Best for
Veo 3.1 Lite	Video + audio	$0.05/sec (720p)	4–8 second clips	High-volume video generation
Lyria 3 Pro	Music	$0.08/song flat	Up to 3 minutes	Full-length soundtrack cues
Grok 4.5	Text	$2 / $6 per Mtok	500K tokens	Widest-context second opinion
Kimi K2.6	Text	$0.95 / $4 per Mtok	256K tokens	Cheapest new frontier-class text tier

Try them now

All four models are live on Ropewalk as of this week — no waitlist, no separate signup. Veo 3.1 Lite is the pick when you need volume over polish: cheaper per second than the full Veo 3.1 tier, with native audio baked into every clip. Lyria 3 Pro is the only music model on the platform that writes a complete 3-minute song rather than a short loop, useful for anything that needs an actual soundtrack rather than a background bed. Between the two text models, Grok 4.5's 500K-token context wins for document-heavy prompts, while Kimi K2.6 is the cheaper of the two and fills a gap — it's the first time any Moonshot AI model has been available on the platform at all.

Every model here passed a real end-to-end generation before shipping in this article: a real rendered video, a real composed song, and two real text replies with token counts and gem costs that matched their listed pricing exactly. That's the same bar the six models in last week's roundup had to clear.

If you're deciding between Ropewalk's text models more broadly, see our ChatGPT vs Claude vs Gemini comparison. For video, our head-to-head of Kling, Hailuo, Wan, and Seedance covers how the existing lineup stacks up against each other.