Scan: Next Grok Model (4.7+) released by September 18, 2026?

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct polymarket_related kalshi_related claude_news article_search gdelt_news wikipedia code_execution

Sub-questions (Fermi decomposition)

What is the current Polymarket YES price and how has it trended since Grok 4.6's release on Aug 12, 2026?
What has been the historical gap (in days) between consecutive Grok point releases (e.g., 4.1 -> 4.2 -> 4.5 -> 4.6), and does a ~37-day gap to Sept 18 fall within that distribution?
Has Elon Musk or xAI publicly stated a target release date for Grok 4.7 or Grok 5, and how reliable have prior Musk timeline claims about Grok releases been?
Are there credible leaks (model identifiers in the Grok app/website, API strings, LMArena anonymous models) indicating a 4.7+ model is imminent?
Is Grok 5 (or another whole-number model) reported to still be in training, and what is its expected completion/release window?
Are there sibling Polymarket or Kalshi markets on Grok 4.7/Grok 5 release dates (other deadlines) that imply a consistent hazard rate?

Planner reasoning

This is a Polymarket question about whether xAI (SpaceXAI) releases a Grok model versioned 4.7 or higher by Sept 18, 2026, given Grok 4.6 shipped Aug 12, 2026. The key drivers are xAI's historical release cadence between point versions, any official roadmap statements (e.g., Grok 5 timelines from Musk), and current leaks/announcements. The market price is the primary anchor, with news search and cadence math as the main supplementary inputs.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.3s	1	## This Market's Polymarket Data Next Grok Model (4.7+) released by September 18, 2026? - Current price (probability): 65.00% - 7-day price change: -20.50% - 30-day price change: -5.00% - Total volume: $15,168 (USD notional) - Price range: 42.00% - 88.50% - Data points: 12 days
polymarket_related	OK	0.9s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'Grok': 0 markets \| keyword 'xAI': 0 markets \| keyword 'Grok 5': 0 markets
kalshi_related	OK	0.8s	0	0 related markets / summaries. keyword 'Grok': no matches \| keyword 'xAI': no matches
claude_news	OK	22.4s	12	## Key Findings (as of Sept 14-15, 2026) - Grok 4.7 still not released. As of Sept 14, 2026, as of September 13, 2026 there is no xAI announcement, model page, API identifier, price or release-notes entry for Grok 4.7, and xAI's newest model is Grok 4.6, released August 12, 2026 . (cellcog.ai)
article_search	OK	0.7s	22	Fetched 22 unique articles across 4 queries (lookback=45d). 'Grok 4.7 release': 10 articles \| 'Grok 5 release date xAI': 10 articles \| 'xAI next Grok model Musk': 10 articles \| 'Grok 4.6 launch xAI': 10 articles
gdelt_news	OK	119.3s	20	GDELT: 20 articles across 3 queries (lookback=30d). 'Grok 5': 10 hits \| 'Grok 4.7': 10 hits \| 'xAI Grok new model': error GDELT rate-limited after retries (429)
wikipedia	OK	0.1s	2	Fetched 2 Wikipedia entries (0 missing pages).
code_execution	OK	35.8s	0	Data setup & assumptions - Known dates: Grok 4 (Jul 9 '25), Grok 4 Fast (Sep 19 '25), Grok 4.1 (Nov 17 '25), Grok 4.6 (Aug 12 '26). Grok 4.2/4.5 dates aren't given, so I distributed the 268-day gap between Grok 4.1 and Grok 4.6 evenly across 3 hypothetical releases (4.2, 4.5, 4.6) → estimated Gr

3. Evidence Brief Sonnet 5 · 7687 chars

# Current state As of Sept 14-15, 2026 (question closes Sept 18, 2026 per title; Kalshi ticker close time is Oct 18, 2026), SpaceXAI's newest publicly shipped model remains **Grok 4.6** (released Aug 12, 2026). No Grok 4.7 (or higher) has been announced, spec'd, or spotted on any leaderboard/API. Musk has pivoted messaging away from "4.7 imminent" toward a new model, **Grok 4.8** (2.5T params, "will finish training this week and start RL" as of Sept 13), with Grok 4.9 and Grok 5 positioned further out on the roadmap with no dates. # Timeline of key events - 2025-11-17 — Grok 4.1 released (confirmed, Wikipedia/research). - 2026-08-12 — Grok 4.6 released, xAI's current top model (confirmed, multiple sources). - 2026-07-24 — Musk says Grok 4.7 "in 4 weeks" (reported, cellcog.ai). - 2026-07-28 — Musk says "a few weeks" after 4.6 (reported). - 2026-08-12 — Musk says "3 to 4 weeks" (reported). - 2026-08-20 — Anonymous "Ox Alpha" model appears on OpenRouter, speculated to be Grok 4.7 (rumored). - 2026-08-26 — Z.ai confirms "Ox Alpha" is actually GLM-5.3-Flash, not Grok 4.7 (confirmed). - 2026-09-01 — Musk says Grok 4.7 "comes out in 10 days" (pointing to ~Sept 11) (reported). - 2026-09-04 — Grok Bot expands to iPad / cheaper plans (confirmed, unrelated to 4.7 model version). - 2026-09-11 — Musk says model "needs a few more days to cook," citing an RL bug re: task-ending behavior (reported). - 2026-09-13 — Musk names successor "Grok 4.8" (2.5T, new C++ stack), says it "will finish training this week and start RL"; no date given (reported). - 2026-09-14 — Musk lays out roadmap: 4.8 → 4.9 (Astra/Fable-class) → Grok 5 ("maybe better than anything"), no dates (reported). - 2026-01-28 — (earlier) xAI Series E announcement confirms Grok 5 in training on Colossus 2 (confirmed). - 2026-08-05 — Musk says Grok 5 "should arrive within 2026" (reported, unconfirmed follow-through). # Event Will SpaceXAI publicly release a Grok model versioned 4.7 or higher by the market's resolution deadline (title says Sept 18, 2026; ticker close Oct 18, 2026)? # Outcomes to forecast Yes / No # Kalshi market anchor No direct Kalshi price returned (kalshi_related found 0 matches for "Grok"/"xAI"). **Polymarket price for this exact market: 65% YES**, but trending sharply down: -20.5% over 7 days, -5% over 30 days, range 42%-88.5% over just 12 data points (low liquidity, $15.2K total volume). The 65% figure appears stale/inconsistent with the latest news (as of Sept 14-15, no 4.7 release, Musk pivoted to 4.8 naming), suggesting further downward correction likely as deadline approaches. # Sub-question answers 1. **Polymarket price/trend since Aug 12 Grok 4.6 release** — Currently 65% YES, down 20.5% in the last 7 days and down 5% over 30 days; hit as high as 88.5% and as low as 42%, reflecting repeated Musk hype-then-slip cycles. (polymarket_direct) 2. **Historical gap between Grok point releases vs. ~37-day window** — Code-execution analysis of gaps (Grok 4→4 Fast→4.1→...→4.6) estimates mean gap ~80 days (std ~14d, low CV); no historical gap has been ≤37 days. Empirical/Weibull fits put P(gap≤37d) near 0-15%; exponential fit overstates at ~37%. Consensus estimate: ~10-15%. (code_execution) 3. **Musk's stated target dates and reliability** — Musk gave five consecutive dated promises for 4.7 (Jul 24, Jul 28, Aug 12, Sep 1, Sep 11), each slipping; as of Sep 13 he abandoned 4.7 messaging entirely in favor of naming a successor, 4.8. Track record: 0-for-5 on self-imposed Grok 4.7 deadlines. (cellcog.ai via claude_news) 4. **Leaks indicating imminent 4.7+** — No credible leak; the one candidate (OpenRouter's "Ox Alpha," Aug 20) was confirmed on Aug 26 to be Z.ai's GLM-5.3-Flash, not Grok. No leaderboard/API sighting of a "Grok 4.7" identifier as of Sept 11. (cellcog.ai) 5. **Grok 5 training status** — Confirmed in training on Colossus 2 as of Jan 28, 2026 Series E announcement; Musk (Aug 5) said "within 2026" but as of Sep 14 roadmap, Grok 5 sits behind both 4.8 and 4.9, with no date given — effectively ruling Grok 5 out for the Sept 18/Oct 18 window. (orcarouter.ai, teslanorth.com) 6. **Sibling markets implying hazard rate** — None found; polymarket_related and kalshi_related both returned 0 matching markets for Grok/xAI/Grok 5. # Key facts (high-confidence, factual) 1. [Wikipedia/research] Grok 4.6 (Aug 12, 2026) is xAI's current released top model; no 4.7+ has shipped as of Sept 14-15, 2026. 2. [cellcog.ai] Musk made five sequential, all-missed release-date claims for Grok 4.7 between Jul 24 and Sep 11, 2026. 3. [cellcog.ai] On Sep 13, 2026, Musk renamed the near-term successor "Grok 4.8," with training completion "this week" and RL to follow — no release date given, and 4.7 was effectively abandoned/superseded. 4. [cellcog.ai] Historical training-finish-to-release gaps for Grok 4.x were 25 days (4.6) and 52 days (4.5); applying this to a mid-Sept 4.8 training finish implies release in early Oct–early Nov, i.e., likely after this market's title deadline (Sep 18) and possibly after the Oct 18 close. 5. [orcarouter.ai] Grok 5 confirmed in training since Jan 2026 Series E; no release date, positioned behind both 4.8 and 4.9 in Musk's Sep 14 roadmap. # Cross-market signals - Kalshi related: none found (0 matches). - Polymarket (own market): 65% YES, but falling fast (-20.5%/7d), thin volume (~$15K), likely lagging the Sept 13-14 news that Grok 4.7 was shelved/renamed 4.8. - Sportsbook implied: N/A. # Analyst opinions and speculation - CellCog.ai and OrcaRouter.ai (AI-tracking blogs) both conclude further slippage is likely given the consistent pattern of missed Musk deadlines and typical 25-52 day training-to-release gaps. - JPMorgan (CNBC, Aug 25) is bullish on Grok monetization broadly but makes no release-date claims. - Musk's own Sept 14 roadmap comments implicitly deprioritize "4.7" as a named milestone, suggesting the next qualifying release will be 4.8 (or later), whose earliest plausible date (per training-to-release analogy) lands in Oct-Nov 2026 — after the title's Sept 18 deadline and possibly after the Oct 18 ticker close. # Directional lean per outcome - **Yes**: Weak support — only the Polymarket 65% legacy price and general "xAI ships fast" reputation; no leaks, no announced imminent release, and Musk has moved on from "4.7" framing entirely. - **No**: Strong support — 0-for-5 missed Musk deadlines, no leaderboard/API evidence, no official 4.7 designation, model renamed to 4.8 with training "this week" (not release), historical gaps (60-90 days) exceed the ~37-day window, and training-to-release analogues (25-52 days) push a plausible release past even the Oct 18 close if the current "4.8, training this week" claim holds. # Gaps / unknowns - Ambiguity between title's Sept 18 deadline and ticker's Oct 18 close time — unclear which governs actual resolution; if Oct 18 governs, probability is meaningfully higher than if Sept 18 governs. - No Kalshi-specific price data was available to anchor against Polymarket's stale 65%. - No confirmation of whether "Grok 4.8" will be branded in a way that satisfies the "4.7+" designation rule (it would, per market rules, since 4.8 > 4.7). # Calibration anchors - Polymarket YES price (anchor): 65%, but trending down sharply (-20.5%/7d) and likely stale relative to Sept 13-14 news. - Historical precedent: 0-for-5 Musk self-imposed Grok deadlines met on time; typical inter-release cadence 60-90 days, all exceeding the ~37-day window to Sept 18. - If Oct 18 close governs: modestly higher odds (~25-35%) given 4.8's "training this week" claim and 25-52 day historical training-to-release lag; if Sept 18 governs: very low odds (~5-10%).

4. Forecasters Fable 5.1 + GPT-5.6 Sol · in parallel

claude-fable-5-1

58% conf

Yes	20%
No	80%

As of Sept 14-15, xAI's latest shipped model is still Grok 4.6 (Aug 12), and Musk has missed five consecutive self-imposed 4.7 dates before pivoting on Sept 13 to a 'Grok 4.8' that is only finishing pretraining 'this week' with RL still to come — historically a 25-52 day lag to release, which lands well past the Sept 18 title deadline. No leaderboard/API sightings of any 4.7+ identifier exist, and the one leak candidate (Ox Alpha) was confirmed to be GLM-5.3-Flash. The Polymarket 65% YES anchor is thin-volume and stale (down 20.5% in 7 days), predating the 4.8 pivot. Residual Yes probability reflects Musk's Sept 11 'few more days to cook' comment on a near-finished model that could still ship as 4.7 within days, plus ambiguity about whether the Oct 18 ticker close rather than Sept 18 governs resolution.

gpt-5.6-sol

72% conf

Yes	12%
No	88%

The only market anchor is Polymarket at 65% YES, but its sharp decline, low liquidity, and apparent failure to incorporate the September 13–14 roadmap news justify a large downward adjustment. Musk missed five successive 4.7 targets, while no credible API, leaderboard, or leak evidence indicates an imminent release. Historical release gaps imply only a 10–15% chance within the relevant short window, and 4.8 is merely finishing training before RL, making release by September 18 highly unlikely. The October 18 ticker close creates some resolution ambiguity and would raise YES odds, preventing an even lower estimate.

5. Devil's Advocate Sonnet 5 · red-team critique

Both forecasts collapse the title/ticker ambiguity into a single number without explicitly stating which deadline they're using, despite the brief noting this is a major swing factor (~5-10% vs ~25-35% depending on resolution date) — neither shows how they weighted the two scenarios or gives a probability split, making their point estimates hard to audit.

Forecast 2's 0.12 Yes seems too low even under the Sept 18 framing, and inconsistent with its own text acknowledging "the October 18 ticker close... would raise YES odds" — if Oct 18 is plausible as the actual resolution date (which the brief flags as ambiguous, possibly the operative one), a blended estimate closer to Forecast 1's 0.20 or higher seems better justified.

Neither forecaster explicitly addresses the base-rate implication from evidence point 4: training-to-release lag of 25-52 days from a "finishing training this week" (~Sept 13-19) claim implies a release window of roughly early-Oct to early-Nov — meaning if Oct 18 governs, the probability could be meaningfully above their ~0.12-0.20, perhaps in the 30-40% range for the back half of that window, yet both anchor near 0.12-0.20 regardless of which close date governs.

Both treat the Polymarket 65% price as legacy/stale and largely discount it, but neither considers that low liquidity ($15.2K volume, only 12 data points) could mean the price hasn't corrected *because* few informed traders have bothered, not necessarily because it's wrong — the "stale" narrative is somewhat speculative rather than confirmed.

Neither forecast considers the possibility that xAI could rush a minor "Grok 4.7" naming/rebrand out of order (e.g., releasing an interim 4.7 checkpoint before 4.8 for competitive/PR reasons) despite Musk's public pivot to "4.8" — companies sometimes ship intermediate patches with lower version numbers than the actively-touted next model, and this scenario of a surprise low-effort 4.7 drop isn't discussed.

Both forecasts lack explicit uncertainty bands or scenario-weighted breakdowns (e.g., P(Yes|Sept18)×P(Sept18 governs) + P(Yes|Oct18)×P(Oct18 governs)), instead presenting single point estimates that obscure how much of the uncertainty is resolution-date-driven versus release-timing-driven.

6. Aggregator Fable 5.1 · final calibrated forecast

17% (65% confidence)

Yes: 17% No: 83%

The title deadline (Sept 18) most likely governs resolution, and under that framing Yes is very unlikely: Grok 4.6 remains the latest shipped model, Musk is 0-for-5 on 4.7 dates, no API/leaderboard sighting exists, and on Sept 13 he pivoted to a 'Grok 4.8' that is only finishing pretraining this week with RL still ahead (P(Yes|Sept 18) ≈ 0.10). The critique is right that the Oct 18 ticker close deserves explicit weighting: with a ~25-52 day training-to-release lag from mid-Sept, a 4.8 release by Oct 18 is plausible but not favored (P(Yes|Oct 18) ≈ 0.35). Weighting ~80/20 toward the title deadline yields roughly 0.15-0.17. The Polymarket 65% is thin, plunging (-20.5% in 7 days), and predates the 4.8 pivot, so I discount it heavily, but the residual chance of a surprise last-minute 4.7 drop (the Sept 11 'few more days to cook' model) keeps Yes from going lower.

Pipeline Timing

Total pipeline time: 218.0s

Per-tool research timings shown in the Research section above.