Two New Models Just Dropped: DeepSeek V4 Pro and Grok 4.6 — LLM Pricing Digest
Weekly roundup from LLM Price Watch — we track live pricing across Claude, GPT, Gemini, DeepSeek, and Grok so you don't have to.
No price cuts or hikes to report this week across our tracked providers. What we did catch: two new model IDs appearing in the wild on August 13th — one from DeepSeek and one from xAI. Here's what we know so far.
DeepSeek V4 Pro (0813)
Model ID: deepseek/deepseek-v4-pro-0813
DeepSeek continues their habit of date-stamping model releases, which is actually a useful practice — it makes it easy to tell at a glance which checkpoint you're running. The "V4 Pro" naming suggests this sits above their previous V3 line, though we don't yet have benchmark comparisons or confirmed context window specs to share. The 0813 suffix matches the date it first appeared in our tracker.
DeepSeek has consistently been one of the more cost-competitive options in our index, so if V4 Pro follows that pattern, it's worth watching. We'll update pricing on the site as soon as it's confirmed.
Practical take: If you're currently running DeepSeek V3 in production and cost-efficiency is a priority, keep an eye on how V4 Pro prices in. DeepSeek's previous generational jumps have held the line on price while improving capability — but don't assume that until the numbers are public.
Grok 4.6
Model ID: x-ai/grok-4.6
xAI is pushing another point release with Grok 4.6, spotted the same day as the DeepSeek drop. Grok 4 was already in our tracker, so this is an incremental update rather than a major new generation. Point releases from xAI have sometimes come with quiet capability improvements or adjusted rate limits, so it's worth checking if you're an active Grok user.
Pricing for 4.6 isn't confirmed in our system yet. Grok 4 has sat at a premium compared to some of the DeepSeek options, so if 4.6 comes in at the same price point, the question is whether the incremental improvements justify staying on that tier versus alternatives.
Practical take: If you're already using Grok 4 via API, test 4.6 on your actual use cases before migrating. Point releases don't always move the needle on the tasks that matter most to your specific workload.
The Bigger Picture This Week
Two new models, zero price changes. That's actually a notable signal in itself. We're in a period where the model release cadence remains fast, but the pricing floor seems to have stabilized — at least for this week. The race to the bottom on token pricing that defined much of early 2025 appears to have leveled off, at least temporarily.
For anyone building on top of these APIs right now, the practical advice is the same as always: don't hard-code model names if you can avoid it, because the deepseek-v4-pro-0813 you integrate today may be superseded by a -0901 variant before your next sprint is done. Abstract your model selection layer where possible.
We'll be watching both of these new models closely over the coming days as pricing details get confirmed and early benchmarks surface.
I track these changes daily at *llmpricewatch.com** — live pricing for Claude, GPT, Gemini, DeepSeek, and Grok, updated automatically. If you want alerts when something changes, the site has you covered.*
Top comments (0)