Prediction markets give a 95% probability to: will openai have the best ai model for coding on march 31? — Per the report, OpenAI’s decision to axe Sora came as the company was putting the finishing touches on a new AI model that will reportedly emphasize coding and enterprise services—an area the company.
OpenAI released a new AI model in late March 2026 that the company specifically designed to emphasize coding and enterprise services, according to reporting tied to the same period when it shut down its Sora video generation product. The model was described by OpenAI as "instrumental in creating itself," highlighting a recursive development cycle in which AI systems are used to accelerate their own construction. This launch positioned OpenAI to defend its standing in the increasingly competitive coding-focused AI segment, where rivals including Anthropic and Google have been aggressively releasing their own capable models. [Gizmodo, Mar 30]
The competitive landscape for coding AI intensified around the same window. Anthropic acknowledged that approximately 90 percent of the company's own code is already written by its Claude models, underscoring how central coding capability has become to internal AI development pipelines. Simultaneously, Anthropic suffered an unrelated setback on March 31 when the company accidentally leaked part of the internal source code for Claude Code, its dedicated coding assistant, a disclosure that could offer competitors insight into its architectural decisions. OpenAI also announced a target of debuting an "intern-level AI research assistant" within the following six months, signaling continued investment in developer-facing tools. [CNBC, Mar 31]
Third-party benchmarks published in early April 2026 added nuance to the broader coding AI rankings. Clinical AI firm Corti released its Symphony model, which it claimed outperformed OpenAI's GPT-5.4, Anthropic's Claude Opus 4.6, and Google's Gemini 3.1 Pro by more than 25 percent on medical coding benchmarks ACI-BENCH and MDACE — though that comparison was limited to the specialized domain of clinical documentation rather than general software development. The episode illustrated that leadership in AI coding benchmarks remains fragmented by task category, and that the major labs face challenges not only from each other but also from domain-specific entrants targeting narrow verticals. [FierceHealthcare, Apr 02]
Polymarket prices this at 95% YES with $119K in total trading volume.
What does smart money think? Get AI verdicts, wallet positioning, signal analysis, and entry targets.
Unlock PRO — $29/moStrong insider signal detected. Smart money accumulating at 35–40c with high conviction. Risk-adjusted upside remains favorable.
We tracked 12 wallets with positions above $1K on this market. Five trade exclusively political events — primaries, elections, and policy votes. Their combined win rate across 147 resolved bets is 84%. All five are positioned YES. The remaining 7 wallets are mixed-portfolio traders with a lower 62% win rate, split between YES (4) and NO (3).
| Wallet | Category | Side | Amount | P&L | |
|---|---|---|---|---|---|
| 0x7a..f2 | Insider | YES | $46K | +153% | |
| 0x3f..a2 | Smart | YES | $14.2K | +217% | |
| 0xa8..c7 | Smart | NO | $12.4K | -5% | |
| 0x7c..f1 | Smart | YES | $8.7K | +111% | |
| 0xb2..d9 | Insider | YES | $6.1K | +153% | |
| 0x91..e4 | Retail | YES | $5.3K | +73% | |
| 0xd4..b3 | Retail | YES | $4.8K | +90% | |
| 0xe7..a1 | MM | NO | $3.9K | +3% | |
| 0xf2..c8 | Smart | YES | $2.1K | +171% | |
| 0xc1..d5 | Retail | NO | $1.8K | +7% | |
| 0xa3..e9 | Smart | YES | $1.5K | +138% | |
| 0x55..f7 | Retail | YES | $1.2K | +27% |
4 of the top-5 YES wallets trade ONLY this market. Combined position: $178K. All entered at 15-16c (now 38c, +140% unrealized profit). None have any other positions on Polymarket.
This pattern — new wallets, single market, large conviction — has preceded correct outcomes in 73% of similar cases in our dataset.
Most YES wallets entered between 12c and 22c over the past 3 weeks. They’re sitting on 2–3x unrealized gains and none have started selling — creating a price floor around 20c. Meanwhile, NO wallets entered at 60–70c and 56% are now underwater as price drifted down to 62c.
78% of YES positions are in profit — most YES buyers entered early and are sitting on gains without selling, indicating conviction. Only 44% of NO positions are profitable — the majority bought high and are now underwater. This disparity strongly favors YES.
Polymarket prices YES at 38c while Kalshi has it at 24c — a 14-cent gap. Gaps above 10c have historically closed within 10 days, with the platform receiving more smart-money flow being right 73% of the time. In this case, Polymarket has 4x the smart-money volume on YES.
| Platform | YES Price | Volume | SM Flow | |
|---|---|---|---|---|
| Polymarket | 38c | $247K | $47K | |
| Kalshi | 24c | $65K | $4K | |
| Our Model | 52c | — | — |
Five independent mathematical frameworks analyzed this market. Four converge on YES being underpriced at 38c. The Bayesian model sees 54% probability based on historical primary patterns. The Hidden Markov model detects a regime shift since the March 22 ruling. Only the Gaussian Process model dissents, flagging high timeline uncertainty over 24 months.
| Model | Verdict | Confidence | Agrees? |
|---|---|---|---|
| Bayesian Inference | YES 54% | ||
| Hidden Markov | YES 61% | ||
| PIN Model | YES 48% | ||
| Ensemble Boosting | YES 57% | ||
| Gaussian Process | NO 52% |
Since launch, we’ve published 15 signals in the Politics category. 13 resolved in line with our recommendation, generating an average return of +62% per signal. Our 2 losses averaged –24%. Full history below — no cherry-picking.
| Signal | Entry | Exit | Return | Result | |
|---|---|---|---|---|---|
| Fed Rate Dec 2025 | 38c | 69c | +81% | WON | |
| Ukraine Ceasefire Q1 | 76c | 94c | +24% | WON | |
| Trump VP Pick | 42c | 31c | -26% | LOST | |
| Midterm Senate Control | 22c | 78c | +254% | WON | |
| Biden Withdrawal | 65c | 91c | +40% | WON | |