LLM Pricing Digest: Two New MiniMax Models and Kimi K3.1 Pro Arrive (Week of Oct 20, 2026)

Weekly pricing intel from LLM Price Watch — tracking Claude, GPT, Gemini, DeepSeek, and Grok.

Quiet week on the price-change front. None of the providers we track moved their rates. No cuts, no increases, nothing to recalculate your cost estimates around.

What did happen: three new models showed up on OpenRouter across two providers. Here's what we spotted and what's worth knowing about each.

MiniMax landed two variants this week. The standard minimax/minimax-m3.1-flash appeared on October 23rd, and a free-tier version — minimax/minimax-m3.1-flash:free — had already shown up a few days earlier on October 20th.

The "flash" naming convention signals this is positioned as a fast, lower-cost option rather than a frontier reasoning model. MiniMax hasn't been a household name in Western developer circles, but they've been quietly building capable models and the free tier is a low-friction way to test throughput and output quality before committing budget.

If you're evaluating cheaper inference options for high-volume tasks — summarization, classification, structured extraction — this is worth a benchmark run. The free variant especially gives you no-cost baseline data to work with.

The other new arrival is moonshotai/kimi-k3.1-pro, first seen October 24th. Moonshot AI (the company behind Kimi) has been building a reputation in the long-context space. The K3 series appears to be their current-generation lineup, and the "Pro" label suggests this sits above a base or flash variant in their hierarchy.

I don't have pricing details to share yet — it's freshly listed and our tracker is still pulling rate data. Worth checking llmpricewatch.com directly if you want the current numbers before they potentially shift as the model matures on the platform.

No price changes sounds boring, but it's useful signal. A few things to take from it:

Stability is good for budgeting. If you're mid-project and have cost estimates locked in around any of the five providers we track, nothing changed this week that would break your math.

New models from smaller providers are worth watching. The competitive pressure that's been driving price cuts over the past year often comes from exactly this pattern: a less-established provider lists something capable at a low price point, and the bigger names eventually respond. MiniMax offering a free flash model is part of that longer dynamic.

Free tiers are useful for evaluation, not just savings. The minimax-m3.1-flash:free variant is most valuable as a zero-cost evaluation lane. Run your actual prompts through it, measure latency and quality against whatever you're currently using, then decide if the paid tier makes sense.

That's the week. No drama, a few new options to experiment with, and steady prices across the board.

For live pricing across Claude, GPT, Gemini, DeepSeek, and Grok — plus whatever new models surface — the tracker runs daily at llmpricewatch.com.