ما الذي تغير في الذكاء الاصطناعي

Proof of work — the maintenance ledger

Every entry below is computed from a dated record — a committed price-refresh file, an approved radar update, or a source-cited model release. Nothing here is typed in by hand.

1. 2026-08-14GLM-5.3 shipped — Z.ai's 14 Aug 2026 release reuses GLM-5.2's base model — every gain comes from scaled post-training on long-horizon RL environments. Large jumps on agentic coding (Terminal-Bench 3.0 4.6 → 28.3, SWE-Marathon 19.4 → 42.5) and an unplanned cyber result (CyberGym 84.5). All figures are Z.ai's own; no independent evaluation exists yet. Weights promised two weeks after launch.
2. 2026-08-12Grok 4.6: same $2/$6 API price as 4.5, AA Intelligence Index 61 — xAI launched Grok 4.6 on 12 Aug 2026 at the Grok 4.5 list price ($2 input / $6 output per 1M tokens under 200k prompt). Official evals: AA Intelligence Index 61 (tied GPT-5.6 Sol Max), Terminal-Bench v3 26%. Migrate off 4.5 with a model-name swap. Not on our LiveBench snapshot yet.
3. 2026-08-12DeepSeek V4 Pro API now serves 0813 weights; promo price unchanged — The deepseek-v4-pro API id now serves DeepSeek-V4-Pro-0813. Promo price stays $0.435 / $0.87 per 1M tokens. DeepSeek's changelog still has no Pro GA note — treat this as a weight drop, not a new product. Flash at $0.14 / $0.28 remains the volume pick.
4. 2026-08-12 — xAI's 12 Aug 2026 successor to Grok 4.5: same $2/$6 API price, same 500K context, a longer post-training run aimed at long-running agents and visual/interactive work. A faster variant costs double. Knowledge cut-off is 1 Feb 2026.
5. 2026-08-12 — The deepseek-v4-pro API id now serves DeepSeek-V4-Pro-0813 weights. Calling method and promo pricing are unchanged. DeepSeek's changelog still has no Pro GA entry — the last note (31 Jul) said an official Pro release would follow soon — so this is a weight drop, not a new product.
6. 2026-08-06 — 21 plans re-checked against the new prices
7. 2026-07-31 — 21 plans re-checked against the new prices
8. 2026-07-31 — ByteDance's new video model generates 30-second audio-and-video clips in one pass, extends them over multiple rounds for multi-minute pieces, and edits by timestamp. At launch it runs inside Jimeng AI and Doubao Pro, with API access announced as coming via BytePlus ModelArk rather than available.