From 5156cd8fab750ca826d37d14a9678fdc3ce97b7b Mon Sep 17 00:00:00 2001 From: Eric Wang Date: Wed, 12 Aug 2026 11:12:46 -0700 Subject: [PATCH] docs: align with DeepSeek-V4-Pro GA (2026-08-12) V4-Pro went GA today as model version DeepSeek-V4-Pro-0813; the model ID stays deepseek-v4-pro and the rate card is unchanged, so the catalog numbers were already correct. What was stale: - catalog.ts provenance comment still described "off-peak" rates and a peak-hour 2x plan upstream has since dropped in favor of an undated overall repricing. Re-verified the rates at GA; the comment now says not to apply the repricing until it lands with an effective date. - README never mentioned the GA. It now notes the GA version, that both models are selectable at every model-choice point, and that pro is the stronger tier for planning/review (the reviewer sub-agent role already runs on it). Thinking-on-by-default and the reasoning-tokens-count-against-max_tokens behavior are called out; no defaults were changed. No runtime behavior changes. bun test tests: 117 pass. Co-Authored-By: Claude Fable 5 --- README.md | 14 ++++++++++++-- src/provider/catalog.ts | 8 +++++--- 2 files changed, 17 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index 21c2e24..4dd49f4 100644 --- a/README.md +++ b/README.md @@ -1,7 +1,9 @@ # deepseek-code (`dsc`) -A DeepSeek-native coding agent for the v4 flash/pro series. TypeScript, -Bun, zero runtime dependencies. +A DeepSeek-native coding agent for the v4 flash/pro series (pro is GA +as of 2026-08-12, model version DeepSeek-V4-Pro-0813 — same +`deepseek-v4-pro` model ID). TypeScript, Bun, zero runtime +dependencies. Not a Claude Code clone and not a general multi-provider harness. Being DeepSeek-only is the design: one protocol state machine (the @@ -56,6 +58,14 @@ Useful flags: --verbose -p only; stream progress to stderr ``` +Both v4 models are 1M context / 384K max output and can be picked +anywhere a model is chosen (`--model`, `/model` in interactive, +`dsc job add --model`). Flash is the default executor; pro is the +stronger tier and the right pick for planning and review passes — the +built-in `reviewer` sub-agent role already runs on it. V4-Pro GA ships +with thinking on by default upstream, and reasoning tokens count +against `max_tokens`, which dsc's budgets already account for. + ## What is actually built - **Provider** (`src/provider/`) — streaming client for the `/anthropic` diff --git a/src/provider/catalog.ts b/src/provider/catalog.ts index 84fa559..e6c5d13 100644 --- a/src/provider/catalog.ts +++ b/src/provider/catalog.ts @@ -1,8 +1,10 @@ // Model catalog with REAL numbers (DESIGN.md: budget math always uses the // measured 616k usable input, never the advertised 1M). -// Pricing: USD per 1M tokens, regular (off-peak) rates, from -// deepseek-docs quick_start/pricing.md (fetched 2026-08-02). A peak-hour -// 2x policy is announced but not yet in effect. +// Pricing: USD per 1M tokens, from deepseek-docs quick_start/pricing.md. +// Re-verified 2026-08-12 at V4-Pro GA (model version DeepSeek-V4-Pro-0813; +// model ID unchanged): rates, context, and max output are unchanged. A +// broad repricing is announced but undated — do NOT apply it here until +// it lands with an effective date. export type ModelInfo = { id: string;