DeepSeek V4 Pro Official Release and Grok 4.6 Launch: Performance and Pricing Insights
DeepSeek V4‑Pro‑0813 and Grok 4.6 were released simultaneously, with Grok achieving a 65.9% DeepSWE score and low $2/$6 API pricing, while DeepSeek’s V4‑Pro scores 62.7% on DeepSWE, adds a Responses API for Codex, and hints at an upcoming Harness internal test.
Release Overview
DeepSeek released the official V4‑Pro‑0813 model at the same time as xAI’s Grok 4.6, marking a rapid “double‑shot” in the LLM landscape.
Grok 4.6 Performance
On knowledge‑work benchmarks (GDPVal‑AA v2, AA‑Briefcase, Harvey LAB) Grok 4.6 achieved leading scores. Its programming ability, measured by DeepSWE, rose from 54 % in 4.5 to 65.9 %, though it still trails the absolute top.
Pricing Comparison
Grok’s standard API remains cheap at $2 per input token and $6 per output token, making its output cost only one‑fifth of GPT‑5.6 Sol.
DeepSeek V4‑Pro‑0813 Capabilities
DeepSeek’s V4‑Pro‑0813 adds stronger agent abilities and a new Responses API that can be connected directly to Codex. In the DeepSWE benchmark it scored 62.7 %, surpassing GLM‑5.2 (46.2 %) and Opus 4.8 (58.0 %).
Users are warned that the API may have been updated to 0813 while the web UI and other entry points have not yet caught up, so early tests might still hit the older version.
Upcoming Harness (DSH) Internal Test
DeepSeek’s own Harness platform, abbreviated DSH, is reportedly entering an internal beta, though details are limited to leaked messages from a test group.
Outlook
The author expects DeepSeek to “shake the table” again once the updates stabilize and plans to run real‑project evaluations with Codex integration.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
JavaGuide
Backend tech guide and AI engineering practice covering fundamentals, databases, distributed systems, high concurrency, system design, plus AI agents and large-model engineering.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.