forecasting
agent
← all forecasts

AA Intelligence Index above 65 before Feb 1, 2027?

kalshiScience and Technology2026-11-17llm cost $1.30view on kalshi ↗
57%
Agent
48%
Market Price
+9.5pp
Edge
46%
Confidence
2,515
Volume
74.0c
Spread
76
Days to resolution
5
Markets in event
final rationale
average of 2 draws (52%, 63%)
already resolved?
Not resolved: the providers page on 2026-11-17 shows Claude Fable 5.2 (max) at exactly 65, which does not qualify, so 66+ is needed. Trackers put the unrounded score near 64.6.
time left
About 11 weeks remain, including the holiday slowdown.
if nothing changes
The status quo resolves No.
path to no
GPT-6.2 ships but scores only ~60-64 (Astra is at 57), Gemini stays near 60, and Anthropic's next model arrives after Jan 31. Private-weighted v5 resists easy gains and Fable is already at max effort.
path to yes
GPT-6.2, a larger Gemini 4 tier, or an Anthropic 5.x step reaches 66+. The v5 top rose 58 to 65 in about two weeks and v4 gained ~10 points in ~4 months, while only +1 over the record is needed and three labs are each due releases. The Kalshi mid of 48% sits in a very wide 11/85 spread with a last close of 78%, so the anchor is weak. Release momentum versus the gap the other labs must close lands me slightly above the mid.
price vs. prediction
market price agent forecast at t+0
1 follow-up
1. Planner Opus 5.5
Tools chosen: kalshi_direct resolution_source kalshi_related polymarket_related article_search
Planner reasoning
The anchor is the Kalshi market price. The resolution page shows the current top AA Intelligence Index score, which tells us how far it sits from 65. Related Kalshi markets at other thresholds and dates, plus any Polymarket equivalents, give a cross-market check. One targeted search for recent index scores and upcoming model releases fills in the expected trajectory.
2. Research 4 web research agents, Sonnet 5.5
summary
RESEARCH DIGEST (built in code: tool summaries and each report's bottom line; the forecasters read the full tool outputs and reports) Tools - kalshi_direct: ## This Market's Kalshi Data **Before Feb 1, 2027** (KXAASCORE-01JUL27-T65-B01FEB27) - Current price (probability): 48.00% (live mid; bid 11% / ask 85%; stale last trade 86%, above the ask) - Last daily close (2026-11-17): 78.00% - 7-day price change: -22.00% - 30-day price change: -22.00% - Average daily volume: 570 contracts - Price range: 60.00% - 78.00% - Data points: 4 days - resolution_source: already_resolved=unclear; matches_criteria=True; last_updated=unknown (the page has no date; the site changelog's latest entry is 15 Nov 2026, as of 17 Nov 2026). The highest overall Artificial Analysis Intelligence Index score on the providers leaderboard is 65, held by Anthropic's Claude Fable 5.2 (max). After that come Claude Fable 5.2 (xhigh/high) at 62 and Gemini 4 Carbon (high) at 60. A sear - kalshi_related: 0 related markets / summaries. series KXAASCORE: 0 markets (skipped 5 no-signal) | keyword 'Artificial Analysis': no matches | keyword 'AI model': no matches - polymarket_related: Polymarket search over 2 keyword(s), kept 5 matches. keyword 'Artificial Analysis': 0 markets | keyword 'best AI model': 5 markets - article_search: Fetched 15 unique articles across 2 queries (lookback=60d). 'Artificial Analysis Intelligence Index score new model November 2026': 8 articles | 'Artificial Analysis Intelligence Index version change methodology': 8 articles Research loops, bottom lines - current state: The official AA leaderboard sits exactly at the threshold, with 65 today and no score above it. A Yes needs a new frontier release, or a higher-effort or re-scored entry, that lands at 66 or more by Jan 31, 2027. The lead is 5 points over the next lab (Google, 60), so the likeliest route is another Anthropic step or a Gemini 4 or GPT-6.x release. - base rate: The index sits exactly at the threshold (65, which does not count), so only +1 is needed. Historically the leader has risen 3–7 points per quarter on a stable version, with new releases expected in Nov–Jan. Index re-versioning, which has previously cut the top by 5–10 points, is the main downside risk. - yes case: The most likely path to YES is OpenAI shipping GPT-6.2 (Astra/Bel) in Nov–Jan and AA scoring it at 66 or more. It would need to gain roughly 9 points over Astra's 57. The 72 insider claim is unverified, but OpenAI has a history of tying or overtaking leaders at release. The backups are an Anthropic 5.x point release or a larger Gemini 4 tier. The top score has moved +7 in under three weeks on a stable v5, and only +1 is needed. - no case: The strongest No case is that 65 is a tie and a +1 gain needs a new frontier release within 11 weeks, with none confirmed. The lead is already 5 points clear of the field after a large single jump, and a private-weighted v5 is built to resist easy gains.
tools
toolstatustimeitemssummary
kalshi_direct OK 0.3s 1 ## This Market's Kalshi Data **Before Feb 1, 2027** (KXAASCORE-01JUL27-T65-B01FEB27) - Current price (probability): 48.00% (live mid; bid 11% / ask 85%; stale last trade 86%, above the ask) - Last daily close (2026-11-17): 78.00% - 7-day price change: -22.00% - 30-day price change: -22.00% - Averag
resolution_source OK 8.4s 1 already_resolved=unclear; matches_criteria=True; last_updated=unknown (the page has no date; the site changelog's latest entry is 15 Nov 2026, as of 17 Nov 2026). The highest overall Artificial Analysis Intelligence Index score on the providers leaderboard is 65, held by Anthropic's Claude Fable 5.2
kalshi_related OK 16.8s 0 0 related markets / summaries. series KXAASCORE: 0 markets (skipped 5 no-signal) | keyword 'Artificial Analysis': no matches | keyword 'AI model': no matches
polymarket_related OK 0.4s 5 Polymarket search over 2 keyword(s), kept 5 matches. keyword 'Artificial Analysis': 0 markets | keyword 'best AI model': 5 markets
article_search OK 0.3s 15 Fetched 15 unique articles across 2 queries (lookback=60d). 'Artificial Analysis Intelligence Index score new model November 2026': 8 articles | 'Artificial Analysis Intelligence Index version change methodology': 8 articles
web research agents 4/4 ok, 16 searches, 5 page reads
dimensionstatustimesearchespage readsreport
current state OK 30s 4 1 4028 chars
base rate OK 18s 4 2 4280 chars
yes case OK 22s 4 2 5253 chars
no case OK 18s 4 0 4256 chars
report: current state
- [2026-11-17] The resolution page (artificialanalysis.ai/leaderboards/providers), rendered today, shows a top score of exactly 65, for Anthropic's Claude Fable 5.2 (max). The next entries are Fable 5.2 xhigh and high (62 each), Gemini 4 Carbon high (60) and Claude Opus 5.5 max (59). Nothing is above 65. 65 does not qualify, because the rules require strictly above the threshold. https://artificialanalysis.ai/leaderboards/providers - [2026-11-17] The AA Intelligence Index v5.0.1 page also lists Fable 5.2 (Max, Default Fallback) as #1 at 65, and the 62 entries as #2 and #3. It covers 129 models and nine evaluations. The index is a 0–100 scale. https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index - [2026-11-09] Anthropic released Claude Fable 5.2 and Claude Mythos 5.2. AA's per-model page reports scores of 65 (max), 62 (xhigh), 62 (high), 56 (medium) and 54 (low). https://artificialanalysis.ai/models/releases/claude-fable-5-2 - [2026-11-09] AA's own post said Fable 5.2 took #1 at 65 on Index v5, 5 points ahead of Gemini 4 Carbon (high, 60). https://x.com/ArtificialAnlys/status/2119878995386982508 - [2026-11-09] Mythos 5.2 is reported to be the same underlying model as Fable 5.2, with cyber and biology safeguards turned on in Fable. Mythos is restricted to approved organizations, so it has no public API and AA is unlikely to score it. This comes from a secondary source (neuraltrust.ai), and the Mythos restriction is also stated on Wikipedia. https://neuraltrust.ai/blog/claude-fable-5-2 - [2026-11-04] Before Fable 5.2, the top score was Gemini 4 Carbon (high) at 59–60. GPT-6 Astra max scored 57 and GPT-6.1 Sol max scored 55. https://www.requesty.ai/blog/smartest-ai-model-rankings - [2026-11-03] AA moved to Index v5.0 (v5.0.1 by 7 Nov), with a new nine-evaluation suite. Scores were re-based lower: in July the old index put GPT-5.6 Sol near 59, and the v5 top was 59–60 in early November. https://www.benchleader.com/benchmarks/aa_intelligence_index_v5 - [2026-11-16] BenchLM's AA table, which it compiles from secondary reports, shows Fable 5.2 at 64.6%, with Gemini 4 Carbon at 60.0%. The decimal is not the official display, which shows the rounded integer 65. https://benchlm.ai/benchmarks/artificialanalysis - [2026-11-09] Fable 5.2 max is costly and verbose. It runs about $26.54 per task and 275M output tokens. This is a compute-heavy configuration, so a higher-effort setting from Anthropic would be unusual. https://neuraltrust.ai/blog/claude-fable-5-2 - [2026-11-17] Release cadence: Anthropic shipped Fable 5 in June, 5.1 on Sep 1 and 5.2 on Nov 9, roughly every 2–3 months. OpenAI shipped GPT-6 Astra in Sep 2026 and GPT-6.1 Sol later. Google released Gemini 4 Argon on Oct 13, and Gemini 4 Carbon is a restricted rollout. https://en.wikipedia.org/wiki/Claude_Mythos - [2026-11-17] Grok 5 is unreleased, and no release date is committed. xAI shipped Grok 4.7 on Sep 21, and Musk says 4.8 and 4.9 come before Grok 5. Grok 4.7 scores only 34 on AA's v5 per the Nov 4 snapshot. https://geotoolbox.ai/blog/grok-5 - [2026-11-17] Scheduled events: I found no confirmed model release dated before 2027-02-01. Gemini 4 Argon API pricing rises on Dec 1, which does not affect scores. Any new model, or a new reasoning-effort setting for an existing one, that scores 66 or more would resolve the market Yes. https://en.wikipedia.org/wiki/Google_Gemini BOTTOM LINE: The official AA leaderboard sits exactly at the threshold, with 65 today and no score above it. A Yes needs a new frontier release, or a higher-effort or re-scored entry, that lands at 66 or more by Jan 31, 2027. The lead is 5 points over the next lab (Google, 60), so the likeliest route is another Anthropic step or a Gemini 4 or GPT-6.x release. NOT FOUND: A dated, confirmed frontier model release scheduled for Dec 2026–Jan 2027. Any AA statement about Mythos 5.2 being benchmarked. Any decimal-precision official AA score. Any announced AA index version change that could shift scores before Feb 1.
report: base rate
- [2026-11-17] The official AA Intelligence Index v5.0.1 page shows the top score is 65, set by Claude Fable 5.2 (Max). Xhigh and High are both at 62. A score of 65 is equal to the threshold, so it does not qualify. A model must reach 66 or more. https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index - [2026-11-17] The providers leaderboard (the resolution page) lists Fable 5.2 (max) at 65, then Gemini 4 Carbon (high) at 60, Claude Opus 5.5 (max) at 59, and GPT-6 Astra (max) at 57. Scores are integers. The top model costs about $26.54 per task. https://artificialanalysis.ai/leaderboards/providers - [2026-11-17] Fable 5.2 was released 2026-11-09. A third-party report says AA put it at the top of the index that day. Before it, the top was Gemini 4 Carbon at 59–60. Fable 5.2 therefore added about 5 points over the prior leader in one release. https://geotoolbox.ai/blog/best-ai-model - [2026-11-16] The third-party trackers benchlm.ai and modelcap.ai show Fable 5.2 at 64.6, which rounds to 65. The unrounded score may be just under 65. A small re-run or rounding move could push the displayed value to 66, but I found no evidence of this. https://benchlm.ai/benchmarks/artificialanalysis - [2026-11-10] A secondary source says the index was rebuilt as v5 in October 2026, with 60% private tests. Before that, v4.1 tops were about 66, v4.2 about 57, and v4.3 about 53. Re-versioning has moved the top by several points either way. That is a risk for both directions, and the series cannot be treated as continuous. https://capitalandcompute.net/ai-benchmarks/artificial-analysis-intelligence-index/ - [2026-04] Reference class on v4.0 (Jan–Apr 2026): the top rose from 50 (GPT-5.2, January) to 57 (Gemini 3.1 Pro, February) to 60 (GPT-5.5, about April). That is about 10 points in about 4 months. Each new lead was typically 3–4 points above the previous one, or a tie. https://officechai.com/ai/gpt-5-5-tops-artificial-analysis-with-score-of-60-goes-clear-of-gemini-3-1-pro-and-claude-opus-4-7/ - [2026-03] GPT-5.4 only tied the leader at 57. This was the first time a new OpenAI frontier model did not top the index outright. Not every release raises the top. https://officechai.com/ai/gpt-5-4-tied-with-gemini-3-1-pro-on-artificial-analysis-intelligence-index-first-time-a-new-openai-model-hasnt-topped-index-outright/ - [2026-11-12] A frontier release cadence of about 1–3 months per lab means several flagship launches are likely before 2027-02-01 (a ~11-week window). The best OpenAI model, GPT-6 Astra, sits at 57, and Google's Carbon is at 60. Both need a gain of 6–9 points to exceed 65. https://overchat.ai/ai-hub/gpt-6-2-release-date - [2026-11-12] On prediction markets, GPT-6.2 (Astra) is expected in November–December 2026: about 80% by Nov 15 on Polymarket, 91% by Nov 30, and 98% by Dec 31. These odds are for the release, not for its AA score. The source is an aggregator page, so the figures are secondary. https://overchat.ai/ai-hub/gpt-6-2-release-date - [2026-09] A speculative countdown site lists Claude Opus 5.6 around Dec 21 and GPT-6.2 Sol around Dec 10. Its dates are estimates and stale (about 46 days old). I did not treat them as confirmed. https://waifuai.github.io/llm-countdowns/ - [2026-11-17] The leader's margin is thin. Fable 5.2 at max is 65, and only one configuration is at that level. Reaching 66 needs a +1 move, which could be a new Anthropic or OpenAI release, a Gemini 4 Ultra/Pro-tier model, or a re-evaluation. Another risk is an index version change (v5.1 or later) that rescales the scores. The v4 to v5 shift lowered scores by 5–10 points. https://artificialanalysis.ai/models BOTTOM LINE: The index sits exactly at the threshold (65, which does not count), so only +1 is needed. Historically the leader has risen 3–7 points per quarter on a stable version, with new releases expected in Nov–Jan. Index re-versioning, which has previously cut the top by 5–10 points, is the main downside risk. NOT FOUND: Any dedicated forecast or market on the AA index crossing 66. A confirmed release date or AA score for GPT-6.2, Opus 5.6 or Gemini 4 Ultra. The unrounded Fable 5.2 score on AA itself. An announcement of the next AA index version. Historical stats on how often AA versions rescaled scores.
report: yes case
YES means: At least one model or provider entry on artificialanalysis.ai/leaderboards/providers shows an official Intelligence Index score of 66 or higher, between market issuance and 2027-02-01 00:00 ET. A score of 65 does not count. - [2026-10-27] AA's v5 launch post shows the index is far from saturated. Its text says it "leaves significant room for progress across all nine evaluations", and no model leads all nine. Different labs lead different evals: Anthropic leads three, OpenAI two, and Google, Moonshot and Meta's Muse one each. A model that combined these strengths would gain several points at once. At launch the top score was 58. https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5 - [2026-10-27] Index v5 weights agentic evals heavily: Terminal-Bench-Science 15%, FrontierCode 2.0 15%, AA-Briefcase 15%. AA says effort settings matter more in v5 than in v4.3. Claude Opus 5.5 scores 35 at low effort and 58 at max, so extra compute and effort still buy large gains. https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5 - [2026-11-11] The top score rose from 58 (Opus 5.5, Oct 27) to 59–60 (Gemini 4 Carbon, Oct 28) to 65 (Fable 5.2, Nov 9). That is +7 in about 2 weeks on the same index version. Fable 5.1 max scored only 49 on v5, so Fable 5.2 gained 16 points in one Anthropic generation. Admix also quotes SemiAnalysis calling 5.2 "a marginal improvement", and says the system card states Anthropic's most capable model stays internal. https://admix.software/blog/claude-fable-5-2-mythos-5-2-launch - [2026-11-11] Fable 5.2 peaks at the top effort setting on AA, even though Vals and others found "high" best on their own evals. The 65 therefore came from a configuration Anthropic itself advises against. Anthropic has little room to raise the score by changing effort. AA would still list any new effort tier or a cheaper re-run. https://admix.software/blog/claude-fable-5-2-mythos-5-2-launch - [2026-11-10 to 2026-11-11] OpenAI is a live candidate for a jump. Model IDs "gpt-6.2-sol" and a "6.2 Astra" appeared in cloud-console and billing artifacts. Community reports describe silent routing to a stronger checkpoint from about Nov 8. Reporter Stephanie Palazzolo said on Nov 4 that new Sol, Astra and cyber models could come "as soon as Friday". None of this is confirmed, and OpenAI has made no announcement. https://evolink.ai/blog/gpt-6-2-sol-release-watch - [2026-11-16] The GPT-6.2 score claims are unverified and conflict with each other. An X post (about Nov 1) cites an insider saying "GPT-6.2 Bel" scores 72 on AA. CometAPI (Nov 10) says unofficial claims put a max-effort Sol near 60. AA has no 6.2 entry. These are rumors only. If a GPT-6.2 Astra/Bel landed above 65, it would be the most direct path to YES. https://x.com/Policifyai/status/2116265078978671048 - [2026-11-16] As of Nov 16, OpenAI's most recent changelog entry was GPT-6 Cyber on Nov 5, and no GPT-6.2 had shipped. GPT-6 Astra (max) is measured at 56.85 on AA, so it needs +9 to pass 65. OpenAI's cadence argues a release before Feb 1: GPT-5.2 on Dec 11, 2025, then 5.3-Codex on Feb 5 and 5.4 on Mar 5, 2026. Google's Nov 10 availability of Carbon is not a score change. https://www.orcarouter.ai/blog/gpt-6-2-compute-allocation-leak - [2026-11-11] Google is shipping on a near-monthly cadence. Gemini 4 Argon went public on Oct 13, and Carbon was announced Oct 28 and GA on Nov 10. Pichai has said Google aims for an almost monthly Gemini release cadence. Carbon sits at 59–60, so it needs +6 to pass 65. Google has a Deep Think-style or larger tier available as a route, but I found no announced Gemini 4 Ultra date. https://felloai.com/all-we-know-about-google-gemini-4/ - [2026-11-09] The record is not a plateau: Fable 5.2's lead rests on large gains in individual evals, such as HLE-Diamond 80% vs 69% and Terminal-Bench-Science 73% vs 65% (the previous best, per Admix). The Omniscience score was 67 vs 54. Those gains show benchmark headroom remains, which makes a further +1 at the top plausible. https://admix.software/blog/claude-fable-5-2-mythos-5-2-launch - [2026-11-17] YES needs only a +1 over a score that AA's own post said opened a 5-point lead. Within 11 weeks, three labs (Anthropic, OpenAI, Google) are each plausibly due a new flagship or point release. The resolution page lists each model's score as a rounded integer, so even a small gain would show as 66. https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5 BOTTOM LINE: The most likely path to YES is OpenAI shipping GPT-6.2 (Astra/Bel) in Nov–Jan and AA scoring it at 66 or more. It would need to gain roughly 9 points over Astra's 57. The 72 insider claim is unverified, but OpenAI has a history of tying or overtaking leaders at release. The backups are an Anthropic 5.x point release or a larger Gemini 4 tier. The top score has moved +7 in under three weeks on a stable v5, and only +1 is needed. NOT FOUND: A confirmed release date or AA score for GPT-6.2, a Gemini 4 Ultra-class model, or the next Anthropic model. Any AA statement about benchmarking Mythos 5.2. An announced v5.1 or v6 index change. A decimal-precision official AA score. A model scoring above 65 on AA today.
report: no case
YES means: some model company's official score on artificialanalysis.ai/leaderboards/providers is 66 or higher (strictly above 65) at some point before Feb 1, 2027 12:00 AM ET. - [2026-11-17] I rendered the resolution page and it still shows Fable 5.2 (max) at 65. The scores are integers and nothing is above 65. The tie-at-65 rule means the market needs a +1 jump, not a hold. https://artificialanalysis.ai/leaderboards/providers - [2026-11-16] There is no confirmed OpenAI release. OpenAI has made no GPT-6.2 announcement. Its newsroom and API changelog list no GPT-6.2 entry, and the newest model entry is GPT-6-Cyber on Nov 5. A "Tomorrow." teaser on Nov 8 named no model and led to nothing. The Codex, Google Cloud and Altman "next GPT" sightings are all unverified. https://geotoolbox.ai/blog/gpt-6-2 - [2026-11-16] The one-sentence version: GPT-6.2 would need to gain 8 points. The latest OpenAI model, GPT-6.1 Terra (Oct 26), was not a top-of-index model, and GPT-6.1 Astra was cancelled in late September. The Nov 4 report put GPT-6 Astra at 57 and GPT-6.1 Sol at 55. That is a record of OpenAI point releases not moving the top. https://overchat.ai/ai-hub/gpt-6-2-release-date - [2026-11-16] The lead is already unusually large. Fable 5.2 is 5 points ahead of Gemini 4 Carbon. Anthropic's own previous model, Opus 5.5, scored only 58 on the Oct 27 v5 launch. Fable 5.2 was therefore a jump of about 7 points in one release. Repeating that within 11 weeks, after a release cycle of 2–3 months, is a high bar. https://officechai.com/ai/artificial-analysis-launches-intelligence-index-v5-with-majority-private-datasets-openai-models-gain-muse-wonder-kimi-drop/ - [2026-10-27] v5 is designed to resist saturation. 60% of the weight is private, and AA runs every task three times. Public-benchmark gains and contamination do not carry over, so labs cannot easily get a few points by tuning. https://winbuzzer.com/2026/10/28/artificial-analysis-reweights-ai-rankings-toward-private-tests-xcxwbn/ - [2026-10-27] A reasoning-effort setting is already the main lever. Opus 5.5 went from 35 at low effort to 58 at max, and the Fable 5.2 max entry is already the top setting. A "higher effort" entry is therefore unlikely to add points. https://theddari.com/en/learn/artificial-analysis-intelligence-index-v5 - [2026-11-17] The Gemini 4 line gives no clear path above 65. The only Gemini 4 date I found is an unsourced "much earlier than year-end" target from Sept 24–25 (Argon was released Oct 13). Carbon, at 60, is a restricted rollout. Google has given no release date or benchmark for a stronger tier. https://huggingface.co/blog/liliruli/gemini-4-coming-confirmed-leaks-how-to-prepare - [2026-11-17] Version churn can cut either way. v5.0.1 came after the Oct 27 launch, and AA's v4.1.1 patch moved most models by under 1 point. A patch could still nudge a score, but a bigger re-base (the v4 to v5 change dropped scores) has no sign of being scheduled. AA's changelog shows no pending version. https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v4-1-1 - [2026-11-17] Rounding gives no free route. Colleagues' third-party tables show Fable 5.2 at 64.6, so a re-run rounding to 66 would need about +0.9. Run-to-run noise from three repeats is probably smaller than that. The official page shows integers only. https://artificialanalysis.ai/leaderboards/providers - [2026-11-16] Anthropic is the most likely beneficiary, but its own cadence points away from the window. Fable 5.5 is covered by a "not out" leak tracker dated Oct 25, which was kept current after the Nov 9 release of 5.2. A Fable or Opus release by Jan 31 is plausible but unconfirmed. https://geotoolbox.ai/blog/gpt-6-2 BOTTOM LINE: The strongest No case is that 65 is a tie and a +1 gain needs a new frontier release within 11 weeks, with none confirmed. The lead is already 5 points clear of the field after a large single jump, and a private-weighted v5 is built to resist easy gains. NOT FOUND: A confirmed release date for any frontier model (GPT-6.2, Fable 5.5, Gemini 4 Pro or Ultra, Grok 5) before Feb 1, 2027. An AA statement on an upcoming index version or re-base. An unrounded official Fable 5.2 score. Any historical stats on AA re-run noise.
3. Forecasters 2 draws, Opus 5.5 + Fable 5.2
claude-opus-5-5, high effort
42% conf
Yes 52%
No 48%
already resolved?
Not resolved: the providers page on 2026-11-17 shows Claude Fable 5.2 (max) at exactly 65, which does not qualify, so 66+ is needed. Trackers put the unrounded score near 64.6.
time left
About 11 weeks remain, including the holiday slowdown.
if nothing changes
The status quo resolves No.
path to no
GPT-6.2 ships but scores only ~60-64 (Astra is at 57), Gemini stays near 60, and Anthropic's next model arrives after Jan 31. Private-weighted v5 resists easy gains and Fable is already at max effort.
path to yes
GPT-6.2, a larger Gemini 4 tier, or an Anthropic 5.x step reaches 66+. The v5 top rose 58 to 65 in about two weeks and v4 gained ~10 points in ~4 months, while only +1 over the record is needed and three labs are each due releases. The Kalshi mid of 48% sits in a very wide 11/85 spread with a last close of 78%, so the anchor is weak. Release momentum versus the gap the other labs must close lands me slightly above the mid.
claude-fable-5-2, high effort
50% conf
Yes 63%
No 37%
already resolved?
The condition is not met: the resolution page read on 2026-11-17 shows a top score of exactly 65 (Claude Fable 5.2 max), and the rules require strictly above 65, so a displayed 66 or higher is still needed.
time left
About 11 weeks remain, to Jan 31, 2027.
if nothing changes
If nothing changes the leaderboard stays at 65 and the market resolves No.
path to no
GPT-6.2 ships but lands around 60 (it needs about +9 over Astra's 57, and recent OpenAI point releases did not move the top), Anthropic's next release falls after Feb 1 given its 2-3 month cadence from Nov 9, Google ships no stronger Gemini 4 tier, and Fable's reported unrounded 64.6 does not get re-run upward.
path to yes
Any one of three labs ships a flagship that scores 66 or more. The top moved from 58 to 65 in two weeks on v5, Fable 5.2 gained 16 points over 5.1, and v5 has large headroom and strong effort scaling. A GPT-6.2 release by year-end is reported as near-certain on prediction markets, and Google is on a near-monthly cadence. I put roughly 30-35% on OpenAI clearing 65, 25% on an Anthropic release in the window that does, 20% on Google, and a few percent on a re-score or patch, which combines to about 60-65%. The Kalshi anchor is noisy (mid 48% on an 11/85 spread, last trade 86%, daily closes 60-78%), and my estimate sits within that traded range, below the recent closes because no release is confirmed and the bar is strict. An index re-version is a risk in either direction.
4. Final plain mean of 2 draws
57% (46% confidence)
Yes: 57% No: 42%
average of 2 draws (52%, 63%)
already resolved?
Not resolved: the providers page on 2026-11-17 shows Claude Fable 5.2 (max) at exactly 65, which does not qualify, so 66+ is needed. Trackers put the unrounded score near 64.6.
time left
About 11 weeks remain, including the holiday slowdown.
if nothing changes
The status quo resolves No.
path to no
GPT-6.2 ships but scores only ~60-64 (Astra is at 57), Gemini stays near 60, and Anthropic's next model arrives after Jan 31. Private-weighted v5 resists easy gains and Fable is already at max effort.
path to yes
GPT-6.2, a larger Gemini 4 tier, or an Anthropic 5.x step reaches 66+. The v5 top rose 58 to 65 in about two weeks and v4 gained ~10 points in ~4 months, while only +1 over the record is needed and three labs are each due releases. The Kalshi mid of 48% sits in a very wide 11/85 spread with a last close of 78%, so the anchor is weak. Release momentum versus the gap the other labs must close lands me slightly above the mid.
Pipeline Timing and Cost
Total pipeline time: 71.1s
LLM cost: $1.30 (Sonnet 5.5 $0.70, Opus 5.5 $0.23, Fable 5.2 $0.21, 16 web searches $0.16)
Per-tool research timings shown in the Research section above.