Revision History

Minor: long-prompt billing sentence cites the model pages; Kimi pricing snapshot dated mid-August; LiveCodeBench wording; GLM-5.1 linked

Minor: LiveCodeBench lead stated only over the models with a reported score

Correction: competitor cells in the frontier comparison table verified against vendor primaries (Kimi K2.6 is 32B active / 1T total and GLM-5.1 40B / 744B, both fully open-weight, not 'Partial'; GPT-5.5 SWE-bench Verified 88.7% had no OpenAI source and is now n/a; Opus 4.6 GPQA ~91% -> 91.3%; filled Opus 4.7, GPT-5.5, Kimi and GLM prices/scores from vendor sources); GPT-5.4/5.5 context 1.05M; benchmark caption no longer says DeepSeek re-ran all competitor figures; captions dated and sourced, refs 78-92 added

Correction: Gemini 3.1 Pro list price in the frontier comparison table was ~$3.50/~$10.50; Google's rate is $2.00/$12.00 (prompts up to 200K). Also: Claude Code mapping sourced and clarified, Pro-0813 description, V4.1-Flash MIT citation, flat-rate dating, thinking-mode rule history

Correction: brought DeepSeek V4 up to DeepSeek's change log as of Sep 23, 2026 (V4-Pro-0813 GA on Aug 13, peak/off-peak pricing from Aug 16, weekend off-peak from Aug 23, V4-Flash and V4-Flash-Vision-Exp retired Sep 10, withdrawn V4-Pro retirement); current pricing table with Aug 1 flat rates kept as history; removed stale 29x-34x comparison; replaced BigGo ref with SCMP; ref 59 section range fixed

Correction: DSec section said the whole platform was Rust-based and that V4's interleaved thinking was tuned through DSec rollouts replacing simulated stubs; the V4 report names three Rust components and does not tie interleaved thinking to DSec. Added the Sep 19, 2026 DSec paper, the V4.1 agent-loop redesign and attributed density figures; repointed dead ref [5] (HF PDF deleted Jun 22, 2026) to arXiv; clarified interleaved-thinking rule.

Auto-saved before update

Auto-saved before update

Cosmetic: wire two previously uncited launch-coverage references into the sentence they support, and drop the MarkTechPost entry whose explanation of the 304B figure the verification pass disproved and which is no longer cited

Correction: fix the Gemini 3.1 Pro price row against Google's published tiers, replace a Terminal-Bench comparison against GPT-5.5 that the cited model card does not contain, remove a pricing-timeline row absent from DeepSeek's change log, correct the DSec spelling, and note that the competitor benchmark columns are DeepSeek's own runs

Correction: swap the Pro and Flash pre-training token counts to match the technical report, remove an AIME 2026 score that appears in neither cited source, correct the checkpoint size to about 865 GB, correct the GPT-5.4 price row against OpenAI's published pricing, and replace the DSpark-module explanation of the 304B figure with the tensor-accounting explanation the Hugging Face data supports

Correction: rebuild the V4-Pro benchmark tables from the model card. The competitor column labelled GPT-5.5 was GPT-5.4 xHigh; its SWE-bench Verified 88.7% is not on the card, its SWE-bench Pro 58.6% was Kimi K2.6's score and its Terminal-Bench 82.7 was GPT-5.4's BrowseComp score. Two Claude Opus 4.6 cells and one MMLU-Pro cell were also wrong, and an unsourced GPT-5.2 column is replaced.

Correction: DeepSeek published open weights for DeepSeek-V4-Flash-0731 on July 31, 2026 (MIT, 48 shards, ~167 GB), so the hub no longer describes the official Flash build as API-only. Added a correctly-framed note that the V4-Pro-DSpark and V4-Flash-DSpark repositories are the same checkpoints with a speculative decoding module attached, confirmed V4-Pro is still the April preview with no official build, re-verified API pricing and the still-inactive peak/off-peak policy against DeepSeek docs, and cross-linked the new DeepSeek V4-Pro page.

Add V4-Flash and V4-Pro variants, Responses API and Codex support, and peak/off-peak API pricing

Added 6 contextual internal links

Add links to related AI Wiki tools

Auto-saved before update

Auto-saved before update

Auto-saved before update

Auto-saved before update

No summary provided