Kalshi AI — Mispricing Audit
Report date 2026-07-07 · Horizon 45 days · Capital $1,000 · Methodology v3 (conviction-weighted, entry-band screened, EV-agnostic sizing) · Picks are live and cleared for auto-copy; rejects are machine-logged and shadow-tracked to settlement.
1. How this was researched
Mode: theme-match. Kalshi has no “AI” category (its categories are Sports, Crypto, Elections, Politics, Economics, Financials, Science and Technology, etc.), so the theme was resolved by keyword sweep over trading_events.title and trading_markets.title across all categories. Terms searched: AI, artificial intelligence, OpenAI, GPT, Anthropic, Claude, Fable, Mythos, Gemini, Grok, xAI, Nvidia, DeepMind, DeepSeek, Llama, Mistral, LLM, chatbot, superintelligence, Sam Altman. Sports-name false positives (tennis players named “Llamas”, “Hanatani”) were curated out by hand. The sweep surfaced ~63 events / 445 markets: five families of NVIDIA GPU rental-price ladders (A100/B200/H100/H200/RTX 5090 — ~330 low-volume strikes), model-release date ladders (GPT-5.6, GPT-6, Gemini 3.5 Pro, Claude Mythos), leaderboard races (LMArena overall/math, Datacurve DeepSWE coding), AI-policy markets (Anthropic’s DoD supply-chain designation, Fable 5 US access), IPO date ladders, and an AI-jobs market.
Stage-1 mechanical screens (24h volume ≥ 1,500; spread ≤ 5¢; not fully priced; not already a published pick; closes within 45 days) cut this to 8 strict passers plus 7 near-misses taken through full diligence. For each survivor: resolution rules pulled from the mirror DB, 14-day intraday history from market_snapshots, live top-of-book from the Kalshi public API, and primary-source news diligence. Two resolution precedents were established from already-settled rungs (queried via the public API): (a) every GPT-5.6 date rung through Jul 7 resolved NO despite the Jun 26 limited-partner release — Kalshi requires true general availability, not a gated preview; (b) KXFABLERESTORE rungs for Jul 7/Jul 10 resolved YES — Fable 5’s US access was restored ~Jul 1, which re-arms the July leaderboard races.
2. Markets reviewed
| Ticker | Question (short) | YES bid/ask | Vol 24h | Disposition |
|---|---|---|---|---|
| KXGPT-OPEN-26JUL09 | GPT-5.6 GA before Jul 9 | 16 / 17 | 11,259 | PICK 1 — BUY NO |
| KXMATHAI-26JUL31-GEMI | Gemini #1 Arena Math, Jul 31 | 28 / 31 | 2,120 | PICK 2 — BUY NO |
| KXTOPMODEL-26JUL31-CLAUF | claude-fable-5 top model, Jul 31 | 8 / 9 | 1,638 | PICK 3 — BUY YES (lottery) |
| KXCHAICUTS-26AUG06-T1 | AI #1 Challenger job-cut reason (July) | 43 / 48 | 131 | Reject — band + volume floor (top watchlist) |
| KXCODEAI-26JUL31-CLAU | Claude top on DeepSWE, Jul 31 | 48 / 49 | 4,605 | Reject — coin-flip band |
| KXCODEAI-26JUL31-CHAT | ChatGPT top on DeepSWE, Jul 31 | 51 / 53 | 1,823 | Reject — coin-flip band |
| KXMATHAI-26JUL31-CLAU | Claude #1 Arena Math, Jul 31 | 56 / 58 | 5,416 | Reject — band screen (edge < 15¢) |
| KXMATHAI-26JUL31-CHAT | ChatGPT #1 Arena Math, Jul 31 | 13 / 14 | 1,680 | Reject — fairly priced |
| KXTOPMODEL-26JUL31-CLAUT | opus-4-6-thinking top model, Jul 31 | 68 / 72 | 387 | Reject — unverifiable resolution view |
| KXGPT-OPEN-26JUL08 | GPT-5.6 GA before Jul 8 | 6 / 8 | 5,168 | Reject — same cluster, thinner edge |
| KXGPT-OPENB-26JUL10 | GPT-5.6 GA before Jul 10 | 75 / 77 | 14,092 | Excluded — already an active pick |
| KXGPT-OPENB-26JUL15 | GPT-5.6 GA before Jul 15 | 84 / 85 | 210 | Reject — no edge, volume died |
| KXGEMINI-GEMI35P-26JUL17 | Gemini 3.5 Pro before Jul 17 | 19 / 21 | 1,233 | Reject — volume floor |
| KXGEMINI-GEMI35P-26JUL24 | Gemini 3.5 Pro before Jul 24 | 67 / 74 | 397 | Reject — 7¢ spread |
| KXANTHROPICRESCIND-26JUN-26AUG01 | DoD/WH rescinds Anthropic designation before Aug 1 | 33 / 34 | 16 | Reject — illiquid |
| KXFRONTIER-FRON-26AUG01 | AI solves a FrontierMath Open Problem before Aug 1 | 15 / 17 | 421 | Reject — volume floor |
| KXFABLEDISABLED-27-26JUL31 | Fable 5 US access disabled again before Jul 31 | 4 / 6 | 94 | Reject — tail, lottery slot used |
Also screened out mechanically: ~330 GPU rental-price ladder strikes (nearly all under the volume floor), Claude Mythos date rungs (1–2¢, fully priced), “Best AI July” brand markets (ANTH at 96–97, fully priced), IPO date ladders (outside horizon or illiquid), and KXTRUMPSAYCOMPANY mention markets (announcer-noise category, skipped by policy).
3. Picks — detail and thesis
Pick 1 — KXGPT-OPEN-26JUL09 · BUY NO @ 0.85 · HIGH CONVICTION
“Will OpenAI release a model called GPT-5.6 or greater before Jul 9, 2026?” — closes 2026-07-09 03:59 UTC (end of Jul 8 ET).
Thesis. For YES to win, OpenAI must take GPT-5.6 to general availability within ~41 hours of this report. Nothing is announced. As of today OpenAI’s own materials describe GPT-5.6 (Sol/Terra/Luna) as a limited preview for ~20 government-vetted partner organizations — “not available in ChatGPT during the preview” — after the U.S. government asked for a staggered rollout on Jun 26. Critically, the resolution precedent is already established: every date rung from Jun 26 through Jul 7 settled NO even though the model was technically shipped to partners, so only a true GA event counts. The residual 16–17¢ on this rung is spillover from the market’s (unconfirmed, self-referential) “GA on Thursday Jul 9” consensus — but a Jul 9 GA resolves this rung NO anyway, since it needs GA on Jul 7–8.
Evidence.- OpenAI, Jun 26: GPT-5.6 preview limited to trusted partners at government request; GA “in the coming weeks” (openai.com; TechCrunch 2026-06-26).
- Kalshi API: KXGPT-OPEN-26JUL07 / JUL06 / JUL05 / OPENB-26JUL03 / JUN30 / JUN26 all finalized NO — limited preview does not count as “release.”
- No GA announcement as of Jul 7 (multiple trackers; OpenAI deployment-safety hub still lists “preview”).
- Price history: this rung fell 60 → 16 over Jul 4–7 as the no-early-GA reality set in; volume 11k/24h accompanied the move (market_snapshots).
Tail risk (cleanest way to lose): OpenAI surprise-ships GA tomorrow (Jul 8) with no pre-announcement — it has a history of sudden drops, and competitive pressure is real (Fable 5 restored Jul 1, Gemini 3.5 Pro rumored Jul 17). I hold 8% for this.
Numbers. 176 contracts × 85¢ = $149.60 at risk (15.0% cap for HIGH tier — sized by tier/band, not EV). Max payout $176.00. Edge +7¢/contract, EV ≈ +8.2% (recorded, not used for sizing).
Liquidity (entry context). YES 16/17; buying NO fills 23 @ 84¢ + 166 @ 85¢ ≈ 189 contracts ≤ limit; vol24h 11,259; OI 13,968. Slippage ≤ 1¢ at this size.
Pick 2 — KXMATHAI-26JUL31-GEMI · BUY NO @ 0.73 · MEDIUM CONVICTION
“Will Gemini be the #1 model on the Arena AI Text Leaderboard (Math) on Jul 31, 2026 at 10:00 AM ET?”
Thesis. On the Jul 1 Arena Math snapshot, Claude holds both top slots — claude-opus-4-6-thinking (1518±12) and claude-fable-5 (1517±37) — while the best Gemini entry is the mid-tier gemini-3.5-flash at 1504±21, fourteen points back. For a YES, Gemini must overtake two entrenched Claude models within 24 days. The realistic path runs through Gemini 3.5 Pro, which (a) is not released, (b) has only an unconfirmed ~Jul 17 launch date after slipping from June, and (c) would then need to accumulate enough Math-category votes for a #1 ranking in under two weeks — Math-slice votes accrue slowly (fable-5, released Jun 9, still has only 243 math votes). The market’s 28¢ on Gemini overweights launch hype relative to that mechanical funnel.
Evidence.- arena.ai Math leaderboard (updated Jul 1): Claude #1 and #2; gemini-3.5-flash #3 at 1504±21 on 815 votes.
- Google announced Gemini 3.5 Pro at I/O (May 19) targeting June GA; slipped to July; still enterprise-preview-only as of Jul 7. Reported Jul 17 date is from secondary sources and unconfirmed (BigGo Finance; Bind AI; TechTimes 2026-06-29).
- Kalshi’s own Gemini ladder agrees on timing: “before Jul 17” trades 19–21¢ while “before Jul 24” trades ~70¢ — launch expected on/after Jul 17, leaving ≤14 days of vote accrual.
- Price history: GEMI crashed 56 → 10 (Jul 2–3), recovered to ~30 on launch-delay chatter — the market is trading hype, not the leaderboard mechanics.
Tail risk (cleanest way to lose): Gemini 3.5 Pro launches ~Jul 17 with its rumored “Deep Think” reasoning layer, debuts as a math monster, and Arena rushes it onto the Math board with a high debut score before month-end. New frontier models have debuted #1 before; that is why this is MEDIUM, not HIGH.
Numbers. 95 contracts × ~73¢ = $69.35 at risk (under the 7% MEDIUM cap). Max payout $95.00. Edge +6¢, EV ≈ +8.2% (recorded, not used for sizing).
Liquidity (entry context). YES 28/31; buying NO fills ~15 @ 72¢ + 176 @ 73¢ ≈ 191 contracts ≤ limit; vol24h 2,120; OI 5,749.
Pick 3 — KXTOPMODEL-26JUL31-CLAUF · BUY YES @ 0.09 · LOW CONVICTION ASYMMETRIC LOTTERY — 2% CAP
“Will claude-fable-5 be the top-ranked AI model on Jul 31, 2026 at 10:00 AM ET?” (Arena leaderboard; ties under Rank (UB) break to highest Arena Score; resolution uses the “Remove Style Control” view.)
Thesis. The market pays 11:1 against the model that is already #1. On the Jul 1 Arena text snapshot, claude-fable-5 sits first at 1509±9 — above claude-opus-4-6-thinking (1504±4) — and the market’s stated tiebreak (highest Arena Score) favors it. Its 4,350-vote count is thin because US access was disabled for most of June; access was restored ~Jul 1 (Kalshi’s own FABLERESTORE rungs settled YES), so July brings a large US vote influx to a model with restoration-hype demand. If its score merely holds through Jul 31, this contract wins at 9¢.
Evidence.- arena.ai overall text leaderboard (Jul 1): claude-fable-5 #1 at 1509±9 (4,350 votes), opus-4-6-thinking #2 at 1504±4 (55,102 votes).
- KXFABLERESTORE-27-26JUL07/JUL10 finalized YES — US access restored ~Jul 1 (Kalshi API).
- Corroborating capability signal: claude-fable-5[max] leads Datacurve DeepSWE at 70%±4 vs gpt-5.5[xhigh] 67%±6 (deepswe.datacurve.ai, Jul 1).
Tail risk (cleanest way to lose) — and why this is a lottery, not a favorite: the market resolves on the style-control-removed view, which I could not retrieve; the persistent 8–9¢ price despite a visible #1 ranking implies informed traders may know fable-5 ranks below opus-4-6-thinking there. Second risk: the post-restore vote flood regresses its thin 1509 downward. Third: GPT-5.6 or Gemini 3.5 Pro debuts #1 before Jul 31, beating every Claude strike at once. Size is capped at 2% of capital accordingly.
Numbers. 222 contracts × 9¢ = $19.98 at risk (2.0% lottery cap — the only sub-35¢ position this run). Max payout $222.00. Edge +16¢, EV ≈ +178% (recorded, explicitly NOT used for sizing — v3 sizes by tier and band only).
Liquidity (entry context). YES 8/9; ~500 contracts offered at 9¢ (NO bid 91 × 500); vol24h 1,638; OI 4,493. Fill is trivial at this size.
4. Recommended $1,000 portfolio
| # | Ticker | Side | Limit | Contracts | Cost | Conviction / band (sets size) | Max payout | Edge ¢ | EV % |
|---|---|---|---|---|---|---|---|---|---|
| 1 | KXGPT-OPEN-26JUL09 | NO | 85 | 176 | $149.60 | HIGH · favorite → 15% cap | $176.00 | +7 | 8.2 |
| 2 | KXMATHAI-26JUL31-GEMI | NO | 73 | 95 | $69.35 | MEDIUM · favorite → 7% cap | $95.00 | +6 | 8.2 |
| 3 | KXTOPMODEL-26JUL31-CLAUF | YES | 9 | 222 | $19.98 | LOW · tail lottery → 2% cap | $222.00 | +16 | 178 |
| Total deployed | $238.93 | Cash held: $761.07 (76.1%) | $493.00 | — | ~22 blended | ||||
Edge and EV columns are recorded for calibration only. They did not set position sizes — the resolved v1/v2 track record showed stated EV was anti-predictive, so v3 sizes strictly by conviction tier and entry band. The blended EV is dominated by the lottery leg, which is precisely why it is capped at 2%.
Cluster exposure (cap: 15% of capital per cluster)
| Cluster | Legs | Cost | % of capital | Cap |
|---|---|---|---|---|
| gpt56-ga-timing | Pick 1 | $149.60 | 15.0% | 15% — at cap |
| lmarena-math-jul31 | Pick 2 | $69.35 | 6.9% | 15% |
| lmarena-top-jul31 | Pick 3 | $19.98 | 2.0% | 15% |
Correlation note: Picks 2 and 3 both ride Arena state on Jul 31 but in opposite failure modes — a Gemini 3.5 Pro mega-debut is the one event that hurts both. Even treated as a single cluster they total $89.33 (8.9%), inside the 15% cap. The GPT-5.6 cluster deliberately holds only the Jul 9 rung; the Jul 8 rung and Jul 15 rung were rejected rather than stacking correlated date legs (the exact v1 failure mode).
Conviction exposure
| Tier | Cost deployed | % of capital | Per-pick cap |
|---|---|---|---|
| HIGH | $149.60 | 15.0% | 15% |
| MEDIUM | $69.35 | 6.9% | 7% |
| LOW / lottery | $19.98 | 2.0% | 2% (tail rule) |
Risk profile
- Worst case: all three legs lose → −$238.93 (−23.9% of capital). Requires a surprise GPT-5.6 GA within 41 hours and a Gemini math takeover and fable-5 losing #1 — largely independent events.
- Best case: all three win → +$254.07 (+25.4%).
- Most likely (per model probs): Picks 1 and 2 win, lottery loses → +$32.07 (+3.2%).
- Concentration: largest single thesis is GPT-5.6 GA timing at exactly the 15% cluster cap; no thesis can flip the run’s sign on its own.
Execution notes
- Limit orders only. Pick 1: rest at 85¢ NO (306 contracts already bid at 83 — join, don’t chase past 85). Pick 2: limit 73¢ NO; book shows ~190 fillable. Pick 3: take the 500-lot offer at 9¢. Kalshi fees (~0.07·p·(1−p)) cost ≈0.9¢, 1.4¢, 0.6¢ per contract respectively — all edges survive.
- Invalidators. Pick 1: any OpenAI GA announcement or scheduled launch event dated Jul 7–8 → exit immediately at market. Pick 2: a confirmed Gemini 3.5 Pro launch before Jul 14 with day-one Arena listing → cut. Pick 3: publication of the style-control-removed ranking showing fable-5 outside the top 2 → the thesis is dead, exit.
- Watchlist triggers. (1) KXCHAICUTS-26AUG06-T1 YES ≤ 48¢ if 24h volume returns above 1,500 — largest modeled edge of the audit (+27¢). (2) KXGEMINI-GEMI35P-26JUL17 NO ~81¢ on a volume print ≥1,500 if the Jul 17 date firms up. (3) KXFRONTIER-FRON-26AUG01 NO ~85¢ on volume.
- Cash is a position: 76% cash reflects an AI board where the liquid coin-flips (DeepSWE race, Math race, GPT-5.6 mid-rungs) are efficiently priced. GA-week volatility around Jul 9–15 should create better entries; the reserve is deliberate.
5. What I rejected and why
Every Stage-2 survivor that didn’t become a pick is machine-logged in picks.json and shadow-tracked to settlement — if the rejects outperform the picks, the screen adds nothing, and we want to know that. Entry-band cuts are marked [band]: this is the v3 discipline making itself visible.
- KXCHAICUTS-26AUG06-T1 (would BUY YES @48, my prob 75%) — [band] + volume floor. AI has been Challenger’s #1 job-cut reason four straight months (14,029 of 45,849 June cuts, 31%), yet July continuation trades at 43–48¢. This is the largest modeled edge of the audit (+27¢) and it is still a reject: a 48¢ entry is the 35–60¢ coin-flip band, which requires HIGH conviction — a future monthly report is not “mechanically locked in” — and 24h volume is 131 vs the 1,500 floor. Exactly the trade v1/v2 would have taken and exactly the profile that lost money. Top watchlist item.
- KXCODEAI-26JUL31-CLAU / -CHAT (48–53¢) — [band]. The DeepSWE coding crown is a genuine coin-flip on (a) GPT-5.6’s GA date and (b) an unknown benchmark score for a model ~20 organizations have touched. CLAU also moved −24¢ in 48h on this exact thesis — the edge, if any, was eaten. Both legs logged.
- KXMATHAI-26JUL31-CLAU (would BUY YES @58, my prob 67%) — [band]. Real lean, +9¢ edge — under the 15¢ coin-flip threshold with only MEDIUM conviction. The same thesis is expressed in the favorites band via NO on the Gemini leg (Pick 2) at better risk shape.
- KXMATHAI-26JUL31-CHAT (NO @87, my prob 85%) — fairly priced; negative edge after fees.
- KXTOPMODEL-26JUL31-CLAUT (68–72¢) — the resolution view (“Remove Style Control”, Rank (UB)) is not publicly fetchable right now; the price implies information about that view I cannot verify. No defensible side, and 24h volume 387 anyway.
- KXGPT-OPEN-26JUL08 (NO @93, my prob 96%) — same cluster as Pick 1 with a thinner residual edge and hours to close; v3 does not stack correlated date rungs.
- KXGPT-OPENB-26JUL10 — already an active pick in the feed (excluded up front; re-emitting would duplicate). Noted for the record: the ~60¢ of probability mass on a Jul 9 GA traces to no primary source I could find.
- KXGPT-OPENB-26JUL15 (85¢) — my estimate ~80% sits below the ask; the 30-day government review window (from Jun 26) could plausibly push GA past Jul 15; 24h volume collapsed to 210.
- KXGEMINI-GEMI35P-26JUL17 (would BUY NO @81, my prob 85%) — below the volume floor (1,233); the NO lean is real (a launch on Jul 17 resolves this rung NO) but the date itself is unconfirmed.
- KXGEMINI-GEMI35P-26JUL24 (67/74) — 7¢ spread fails the 5¢ screen; thesis would rest on an unconfirmed date.
- KXANTHROPICRESCIND-26JUN-26AUG01 (33/34) — 16 contracts of 24h volume; unfillable. Directionally interesting (preliminary injunction Mar 26, WH meeting Apr 17, Fable 5 restored Jul 1 all point to thaw) but no primary-source rescission action exists, and 34¢ is the tail band.
- KXFRONTIER-FRON-26AUG01 (would BUY NO @85, my prob 92%) — no FrontierMath Open Problem has ever been solved and the price has drifted down all week, but 421 of 24h volume fails the floor.
- KXFABLEDISABLED-27-26JUL31 (4/6¢) — [band] sub-35¢ tail; the single lottery slot went to CLAUF; re-escalation risk at ~5¢ looks roughly fair anyway.
6. Sources
- TechCrunch — OpenAI limits GPT-5.6 rollout after government request (Jun 26, 2026)
- OpenAI — Previewing GPT-5.6 Sol
- OpenAI Deployment Safety Hub — GPT-5.6 Preview System Card
- Constellation Research — GPT-5.6 limited preview pending US government sign-off
- TechTimes (Jul 6, 2026) — GPT-5.6 release nears
- arena.ai — Text leaderboard (updated Jul 1, 2026) and Math category
- Datacurve DeepSWE leaderboard (v1.1, updated Jul 1, 2026)
- TechTimes (Jun 29, 2026) — Gemini 3.5 Pro cleared for July launch; Fable 5 nears return
- BigGo Finance — Google delays Gemini 3.5 Pro launch to July 17 (unconfirmed date; treated as secondary)
- Bind AI — Gemini 3.5 Pro slips to July
- Challenger, Gray & Christmas — June 2026 report: AI leads job-cut reasons for fourth consecutive month
- Mayer Brown — Anthropic supply-chain-risk designation: latest developments
- Pearl Cohen — Anthropic sues DoD over supply-chain-risk designation
- Kalshi public API (api.elections.kalshi.com/trade-api/v2) — live orderbooks and finalized results for KXGPT-OPEN*/KXGPT-OPENB* (all NO through Jul 7) and KXFABLERESTORE-27-26JUL07/JUL10 (YES).
- Kalshi mirror DB (market_snapshots, trading_markets, trading_events) — rules text, 14-day intraday price/volume history.