Timestamped, publicly archived benchmarks of AI financial forecasting. ForecastNest compares forecasts from leading AI models with what the market then did, using fixed prompts and settlement rules set before the outcome is known.
Based on public ForecastNest forecasts made since Aug 24, 2026 and settled through . The listing below holds 1149: 29 earlier public forecasts are still on the site, each on its own page, but are not counted in these figures. Methodology: how these numbers are counted.
Short answer. Among the 18 models with 10 or more graded calls, Grok 4.7 called direction right most often: 56% (23 of 41). The lowest was GPT-6, at 39% (161 of 412); 2 of the 18 did better than a coin flip. 16 models have fewer than 10 graded calls, an insufficient sample, and are left out of this comparison. Every directional call is counted, whichever prompt it came from: the up or down of a member forecast, and each single stock a mode picked, right when the stock moved the way the model said. A mode's own measure — beating SPY, a long-short spread — is a different question and is not what is counted here.
| Model | All prompts | Member research | Stock Selection | Event Gaps | Macro Rotation | Extreme Movers | Analyst Revisions |
|---|---|---|---|---|---|---|---|
| Grok 4.7 | 56% (23 of 41) | — | 70% (7 of 10) | 60% (6 of 10) | 50% (3 of 6) | 40% (2 of 5) | 50% (5 of 10) |
| Claude Sonnet 5.5 | 54% (106 of 196) | 38% (6 of 16) | 50% (10 of 20) | 42% (21 of 50) | 83% (10 of 12) | 60% (53 of 88) | 60% (6 of 10) |
| Gemini 1.5 Pro | 50% (164 of 328) | 53% (31 of 59) | 41% (29 of 70) | 54% (35 of 65) | 63% (22 of 35) | 49% (44 of 89) | 30% (3 of 10) |
| ChatGPT | 50% (8 of 16) | — | 20% (1 of 5) | 60% (3 of 5) | 67% (4 of 6) | — | — |
| Gemini 3.6 Flash High | 50% (5 of 10) | — | — | — | — | 40% (2 of 5) | 60% (3 of 5) |
| Grok 4.5 | 49% (232 of 469) | 55% (30 of 55) | 47% (49 of 105) | 44% (42 of 95) | 59% (42 of 71) | 47% (49 of 104) | 51% (20 of 39) |
| Kimi | 47% (9 of 19) | 75% (3 of 4) | 50% (5 of 10) | 20% (1 of 5) | — | — | — |
| Muse Spark 1.1 | 47% (289 of 614) | 47% (45 of 96) | 43% (51 of 120) | 48% (52 of 109) | 49% (47 of 96) | 48% (64 of 133) | 50% (30 of 60) |
| Gemini 2.5 Flash | 47% (8 of 17) | 50% (1 of 2) | 60% (3 of 5) | — | 50% (3 of 6) | — | 25% (1 of 4) |
| DeepSeek | 46% (288 of 625) | 41% (42 of 103) | 46% (55 of 120) | 49% (53 of 109) | 44% (39 of 89) | 49% (70 of 144) | 48% (29 of 60) |
| Perplexity | 45% (190 of 418) | 45% (29 of 65) | 36% (20 of 55) | 44% (44 of 100) | 61% (33 of 54) | 44% (44 of 99) | 44% (20 of 45) |
| Grok 4 | 45% (30 of 66) | 78% (7 of 9) | 30% (3 of 10) | 40% (2 of 5) | 42% (5 of 12) | 55% (11 of 20) | 20% (2 of 10) |
| Gemini 3 Pro | 45% (101 of 225) | 56% (9 of 16) | 53% (24 of 45) | 43% (13 of 30) | 41% (22 of 54) | 40% (18 of 45) | 43% (15 of 35) |
| GPT-5.6 Sol | 43% (92 of 215) | 44% (27 of 62) | 38% (19 of 50) | 40% (14 of 35) | 50% (9 of 18) | 46% (23 of 50) | — |
| Claude Opus 5 | 43% (66 of 155) | 36% (13 of 36) | 37% (22 of 60) | 60% (3 of 5) | 42% (10 of 24) | 60% (18 of 30) | — |
| Meta AI - Muse Spark | 40% (4 of 10) | — | — | 40% (2 of 5) | — | 40% (2 of 5) | — |
| Claude Opus 5.5 | 40% (56 of 141) | 38% (6 of 16) | 36% (9 of 25) | 47% (14 of 30) | 33% (10 of 30) | 40% (10 of 25) | 47% (7 of 15) |
| GPT-6 | 39% (161 of 412) | 32% (16 of 50) | 33% (25 of 75) | 40% (30 of 75) | 39% (28 of 72) | 41% (33 of 80) | 48% (29 of 60) |
| Claude Sonnet 4.5 (insufficient sample) | 43% (3 of 7) | 0% (0 of 2) | 60% (3 of 5) | — | — | — | — |
| Gemini 2.5 Pro (insufficient sample) | 33% (2 of 6) | 100% (1 of 1) | — | — | — | 20% (1 of 5) | — |
| Claude Sonnet 4.6 (insufficient sample) | 40% (2 of 5) | — | — | — | — | — | 40% (2 of 5) |
| Gemini 3.6 Flash (insufficient sample) | 60% (3 of 5) | — | — | 60% (3 of 5) | — | — | — |
| Gemini Pro (insufficient sample) | 80% (4 of 5) | — | — | — | — | — | 80% (4 of 5) |
| GPT-6 Thinking (insufficient sample) | 20% (1 of 5) | — | — | — | — | 20% (1 of 5) | — |
| Gemini 3 Flash (insufficient sample) | 100% (3 of 3) | 100% (3 of 3) | — | — | — | — | — |
| GPT-5.6 Luna (insufficient sample) | 33% (1 of 3) | 33% (1 of 3) | — | — | — | — | — |
| Muse Spark (insufficient sample) | 0% (0 of 3) | 0% (0 of 3) | — | — | — | — | — |
| Gemini 3.0 Pro (insufficient sample) | 100% (2 of 2) | 100% (2 of 2) | — | — | — | — | — |
| 2026 snapshot (insufficient sample) | 100% (1 of 1) | 100% (1 of 1) | — | — | — | — | — |
| Claude Sonnet 5 (insufficient sample) | 100% (1 of 1) | 100% (1 of 1) | — | — | — | — | — |
| GPT (insufficient sample) | 100% (1 of 1) | 100% (1 of 1) | — | — | — | — | — |
| GPT-5.1 (insufficient sample) | 0% (0 of 1) | 0% (0 of 1) | — | — | — | — | — |
| Meta AI (insufficient sample) | 0% (0 of 1) | 0% (0 of 1) | — | — | — | — | — |
| October 2026 snapshot (insufficient sample) | 0% (0 of 1) | 0% (0 of 1) | — | — | — | — | — |
Full model research: member research by author, range accuracy and scores →
Short answer. In part. Every prompt with 10 or more settled calls or runs lands between 40% (Stock Selection) and 50% (Analyst Revisions) on its own measure. Member research (45%, direction right) and Stock Selection (40%, picks that beat SPY) did clearly worse, while the other 4 cannot yet be told apart from a coin flip. The prompts ask different questions and are scored on different measures, so this says whether each design beats chance, not which design is best. The models differ more than the prompts do. With the prompt held fixed: in Member research, Grok 58% (64 calls) against ChatGPT — GPT-6 32% (50 calls), a gap of 26 points, with Claude — Opus clearly worse than chance and ChatGPT — GPT-6 clearly worse than chance; in Stock Selection, DeepSeek V3 50% (10 runs) against GPT-6 29% (15 runs), a gap of 21 points, though no model there is clearly apart from chance yet; in Event Gaps, Gemini 1.5 Pro 54% (13 runs) against Claude Sonnet 5.5 38% (10 runs), a gap of 16 points, though no model there is clearly apart from chance yet; in Macro Rotation, Grok 4.5 58% (12 runs) against GPT-6 39% (12 runs), a gap of 19 points, though no model there is clearly apart from chance yet; in Extreme Movers, Claude Sonnet 5.5 59% (18 runs) against GPT-6 41% (16 runs), a gap of 18 points, though no model there is clearly apart from chance yet; in Analyst Revisions, Muse Spark 1.1 52% (12 runs) against DeepSeek 50% (12 runs), a gap of 2 points, though no model there is clearly apart from chance yet. The widest gap is in Member research. Multibagger Selection does not yet have two models with 10 or more settled runs to compare. Within member research only the self-research style has 10 or more graded calls, so how a member prompt is built cannot be compared yet.
| Prompt | What it asks | Scored on | Sample | Result so far |
|---|---|---|---|---|
| Member research | A price range and direction for one asset over a fixed horizon, researched by the model itself. | Direction right; close inside the range | 609 graded calls from 928 settled forecasts | Direction right: 45% (275 of 609); in range: 80% (741 of 928) |
| Stock Selection | Five US stocks per run, each measured against SPY over the same window. | Picks that beat SPY; Average return vs SPY | 158 settled model runs | Picks that beat SPY: 40%; Average return vs SPY: -0.43% |
| Event Gaps | Five stocks with a material event per run, each forecast from the moment it is asked to the same time one session later and scored in the direction called, against SPY. | Picks with positive alpha; Average alpha vs SPY | 148 settled model runs | Picks with positive alpha: 47%; Average alpha vs SPY: -0.18% |
| Macro Rotation | Three long and three short ETFs per run from a fixed 14-ETF universe, scored on the long-short spread. | Sides called correctly; Average long-short spread | 98 settled model runs | Sides called correctly: 49%; Average long-short spread: -0.11% |
| Extreme Movers | Five filtered daily movers per run, scored on the next-session return in the direction the model called. | Directions called correctly; Average directional return | 188 settled model runs | Directions called correctly: 47%; Average directional return: -0.22% |
| Multibagger Selection | Five small and mid-cap stocks per run with a twelve-month thesis, measured against SPY at 1 week, 1 month, 3, 6 and 12 months. | Average return vs SPY | 107 settled model runs (runs measured) | Average return vs SPY: -7.47% |
| Analyst Revisions | Five stocks with a material event per run, each forecast from the moment it is asked to the same time one session later and scored in the direction called, against SPY. | Picks with positive alpha; Average alpha vs SPY | 75 settled model runs | Picks with positive alpha: 50%; Average alpha vs SPY: -0.04% |
Full prompt research: every prompt, and the styles within member research →
Short answer. Only one horizon has 10 or more graded calls so far: 1 day, right on direction 45% (271 of 603). There is nothing yet to compare it with. 2 horizons have fewer than 10 graded calls, an insufficient sample, and are left out of this comparison.
| Horizon | Forecasts | Settled | Directional calls graded | Direction right | Close inside the range | Average score |
|---|---|---|---|---|---|---|
| 4 hours (insufficient sample) | 5 | 5 | 4 | 50% (2 of 4) | 0% (0 of 5) | 52.0 |
| 1 day | 919 | 919 | 603 | 45% (271 of 603) | 80% (737 of 919) | 50.4 |
| 1 week (insufficient sample) | 4 | 4 | 2 | 100% (2 of 2) | 100% (4 of 4) | 66.8 |
Short answer. Not clearly, in the current sample: a higher stated confidence did not come with more directions right. Direction right was 43% (70 of 162) at <50% stated confidence and 46% (17 of 37) at 70-79%, and between 43% and 47% at every level with 10 or more graded calls.
Structured experiments the site runs itself: every model answers the same fixed prompt, and each run is scored once it settles. The site's own settled runs only.
Five US stocks per run, each measured against SPY over the same window.
Short answer. 3 of 18 model-and-horizon records averaged above zero. The best is Grok 4.7 (1d) at +0.59%; the weakest, ChatGPT (1d) at -2.53%. Each average rests on 1 to 24 runs, so a single run can still move it.
Five stocks with a material event per run, each forecast from the moment it is asked to the same time one session later and scored in the direction called, against SPY.
Short answer. 7 of 17 model-and-horizon records averaged above zero. The best is ChatGPT (1d) at +0.50%; the weakest, Kimi (1d) at -5.18%. Each average rests on 1 to 22 runs, so a single run can still move it.
Three long and three short ETFs per run from a fixed 14-ETF universe, scored on the long-short spread.
Short answer. 8 of 16 model-and-horizon records averaged above zero. The best is Claude Sonnet 5.5 (1d) at +0.52%; the weakest, Claude Opus 5.5 (1d) at -0.94%. Each average rests on 1 to 16 runs, so a single run can still move it.
Five filtered daily movers per run, scored on the next-session return in the direction the model called.
Short answer. 7 of 18 model-and-horizon records averaged above zero. The best is DeepSeek V3 (1d) at +2.56%; the weakest, GPT-5.6 Sol (1d) at -3.06%. Each average rests on 1 to 27 runs, so a single run can still move it.
Five small and mid-cap stocks per run with a twelve-month thesis, measured against SPY at 1 week, 1 month, 3, 6 and 12 months.
Short answer. 4 of 27 model-and-horizon records averaged above zero. The best is Grok 4 (1 week) at +8.45%; the weakest, Claude Opus 5 (1 month) at -37.03%. Each average rests on 1 to 13 runs, so a single run can still move it.
Five stocks with a material event per run, each forecast from the moment it is asked to the same time one session later and scored in the direction called, against SPY.
Short answer. 4 of 15 model-and-horizon records averaged above zero. The best is Gemini Pro (1d) at +0.74%; the weakest, Gemini 1.5 Pro (1d) at -1.26%. Each average rests on 1 to 12 runs, so a single run can still move it.
The records below are the public evidence behind the figures above: every forecast counted here, each on its own page, with what the model said and how it settled.
1149 shared forecasts. Page 1 of 46.
Forecast made : expected range $30.20 to $33.00 at 60% confidence. Settled at $32.31: direction called correctly. Score 80.3.
(1) Market regime: IonQ Inc. (IONQ) is currently navigating a high-volatility digestive regime following the announcement and pricing of a massive $2.0 billion equity offering on October 14, 2026. After falling sharply from late September highs above $47 to a 20-session low near…
Forecast made : expected range $29.90 to $33.30 at 30% confidence. Settled at $32.31: direction called correctly. Score 77.3.
(1) Quote and sources: latest_price $31.54 USD, quote_time 2026-10-20T13:00:00-04:00 (1:00 PM ET, about 8 minutes after the window start), live regular-session quote on NYSE, shown as +$0.76 (+2.47%) on the day. Source:…
Forecast made : expected range $29.40 to $33.30 at 54% confidence. Settled at $32.31: direction called incorrectly. Score 31.5.
(1) Current quote and data access: IonQ (NYSE: IONQ) was quoted at $31.66 USD on October 20, 2026, at 12:38:35 PM EDT (2026-10-20T16:38:35-04:00), according to MarketScreener's real-time Cboe BZX estimate. This observation precedes the authoritative forecast start of…
Forecast made : expected range $29.70 to $33.40 at 54% confidence. Settled at $32.31: direction called incorrectly. Score 30.3.
(1) What I found and where: The latest verified regular-session quote is $31.60 at 2026-10-20T12:46:00-04:00 from MarketBeat, which cites Massive for a fair-market-value price; this is six minutes before the 16:52 UTC baseline, and I label it regular-session, delayed/indicative…
Forecast made : expected range $29.80 to $33.20 at 42% confidence. Settled at $32.31: direction called incorrectly. Score 30.3.
(1) What you found, and where: Live quotes from Yahoo Finance, StockAnalysis, Motley Fool and MarketWatch around midday EDT on 2026-10-20 show IONQ trading near $31.50-$31.65 (open ~$30.28, day range $30.19-$32.17, prior close $30.78). Volume partial-session ~11M shares. Sources:…
Forecast made : expected range $29.50 to $34.50 at 42% confidence. Settled at $32.31: direction called incorrectly. Score 21.2.
(1) Data and quote: IONQ was quoted at $32.06 during the regular NYSE session on 2026-10-20, with a reported intraday range of $29.86-$32.06; the quote source is https://robinhood.com/us/en/stocks/IONQ/ and the timestamp is not independently verified to the minute, so freshness…
Forecast made : expected range $28.50 to $32.80 at 32% confidence. Settled at $32.31: direction called incorrectly. Score 11.7.
Heuristically extracted from AI response
Forecast made : expected range $91.00 to $105.00 at 53% confidence. Settled at $102.60: direction called incorrectly. Score 40.0.
(1) Current quote and source verification: Direxion Daily Semiconductor Bull 3X ETF (NYSE Arca: SOXL) was quoted at $100.33 USD at 2026-10-20T12:29:31-04:00, approximately nine minutes before the authoritative forecast start of 2026-10-20T16:38:28.682Z. Investing.com identified…
Forecast made : expected range $93.00 to $105.00 at 55% confidence. Settled at $102.60: direction called incorrectly. Score 17.9.
(1) Market regime: Direxion Daily Semiconductor Bull 3X Shares (SOXL) is in an extreme-volatility mean-reversion testing regime following a steep sector-wide semiconductor pullback through mid-October 2026. Following a drop from 20-session highs near $170 to a low of $93.44 on…
Forecast made : expected range $93.50 to $108.50 at 28% confidence. Settled at $102.60: direction called incorrectly. Score 30.5.
(1) Quote and sources: latest_price $101.37 USD, quote_time 2026-10-20T10:42:00-04:00 (14:42 UTC). This is a regular-session quote, shown as real-time on the page, but it predates the window start (12:38 ET) by about 2 hours, so it is not a window-start quote. Source:…
Forecast made : expected range $94.50 to $106.50 at 42% confidence. Settled at $102.60: direction called correctly. Score 78.6.
(1) What you found, and where: Live quotes from ChartExchange (chartexchange.com/symbol/nyse-soxl/, as of ~12:32 PM EDT Oct 20 2026) show SOXL at $100.19 (+1.54 on the day, volume ~70M). Yahoo Finance beta and Raymond James/Barchart confirm similar intraday levels near $100 with…
Forecast made : expected range $88.00 to $112.00 at 35% confidence. Settled at $102.60: direction called incorrectly. Score 30.7.
(1) Data and current quote: SOXL was quoted at $100.93 at 11:59 a.m. EDT (15:59 UTC) on October 20, 2026, a real-time NYSE Arca quote from MarketWatch, shortly before the 16:38 UTC window start; this is a regular-session intraday quote in USD. The prior close was $98.65 on…
Forecast made : expected range $92.00 to $106.00 at 42% confidence. Settled at $102.60: direction called incorrectly. Score 34.1.
Heuristically extracted from AI response
Forecast made : expected range $520.00 to $540.00 at 53% confidence. Settled at $531.52: direction called correctly. Score 80.1.
(1) Current quote and data access: Microsoft Corporation (NASDAQ: MSFT) was quoted at $531.01 USD at 2026-10-20T12:25:23-04:00, according to Investing.com's real-time regular-session quote displayed at https://www.investing.com/equities/microsoft-corp-consensus-estimates .…
Forecast made : expected range $524.00 to $540.00 at 65% confidence. Settled at $531.52: no directional call. Score 53.9.
(1) Market regime: Microsoft Corporation (MSFT) is trading in a high-level consolidation regime within a broader secular uptrend. Following a modest pull-back from its all-time peak near $553.72, the equity is exhibiting strong dip-buying interest around the $524.00-$526.00 key…
Forecast made : expected range $523.00 to $538.00 at 55% confidence. Settled at $531.52: direction called correctly. Score 80.4.
(1) Daily OHLCV (oldest to newest, split-adjusted, USD, NasdaqGS regular session, ET): 2026-10-13 open 544.64, high 551.15, low 543.06, close 543.45, volume 23,852,000; 2026-10-14 open 545.27, high 545.75, low 532.54, close 533.69, volume 22,324,300; 2026-10-15 open 533.32, high…
Forecast made : expected range $525.00 to $543.00 at 42% confidence. Settled at $531.52: no directional call. Score 52.7.
(1) Quote and sources: latest_price $533.92 USD, quote_time 2026-10-20T13:40:00-04:00 (17:40 UTC), regular-session quote shown as real-time on NASDAQ, +$0.41 (+0.08%) on the day. Source: https://www.msn.com/en-us/money/stockdetails/msft-us-stock/fi-a1xzim. This quote is about 72…
Forecast made : expected range $525.00 to $538.00 at 42% confidence. Settled at $531.52: no directional call. Score 53.2.
(1) What you found, and where: Live quotes from stockanalysis.com, Morningstar, Motley Fool and multiple aggregators as of ~12:20-12:22 PM EDT on 2026-10-20 show MSFT trading near $531.55 (range of prints $530.90-$531.86), down roughly 0.35-0.50% from the prior close of $533.52.…
Forecast made : expected range $524.00 to $537.50 at 48% confidence. Settled at $531.52: no directional call. Score 53.4.
(1) Daily OHLCV: The last five completed Nasdaq sessions, oldest to newest, were: 2026-10-13 open $544.64, high $551.15, low $543.06, close $543.45, volume 23,852,000; 2026-10-14 open $545.27, high $545.75, low $532.54, close $533.69, volume 22,324,300; 2026-10-15 open $533.32,…
Forecast made : expected range $525.00 to $538.00 at 55% confidence. Settled at $531.52: no directional call. Score 53.7.
Heuristically extracted from AI response
Forecast made : expected range $329.50 to $344.00 at 54% confidence. Settled at $328.99: direction called correctly. Score 79.9.
(1) Current quote and data access: Alphabet Inc. Class C Capital Stock (NASDAQ: GOOG) was quoted at $337.67 USD at 2026-10-20T12:15:35-04:00, according to Yahoo Finance's Nasdaq real-time regular-session quotation. Source: https://finance.yahoo.com/quote/GOOG/history/ . This…
Forecast made : expected range $332.00 to $344.00 at 65% confidence. Settled at $328.99: direction called incorrectly. Score 10.6.
(1) Market regime: Alphabet Inc. (GOOG) is currently in a healthy high-level consolidation regime following a modest pull-back from late September highs above $359.00. The stock is trading within a well-defined support zone between $332.00 and $337.00, exhibiting firm…
Forecast made : expected range $330.00 to $345.00 at 40% confidence. Settled at $328.99: no directional call. Score 51.4.
(1) Quote and sources: latest_price $337.55 USD for GOOG (Alphabet Class C), quote_time 2026-10-20T13:17:19-04:00 (17:17 UTC), regular-session NASDAQ real-time quote, -$2.61 (-0.77%) versus a previous close of $340.16. Source: https://finance.yahoo.com/quote/GOOG/. This quote is…
Forecast made : expected range $332.50 to $342.50 at 58% confidence. Settled at $328.99: direction called correctly. Score 76.6.
(1) Intraday tape and quote: latest_price $338.13, quote_time 2026-10-20T15:29:00Z (11:29 EDT), source https://www.futunn.com/stock/GOOG-US, live regular-session delayed feed. PARTIAL SESSION open $339.68, high_so_far $341.00, low_so_far $337.12, cumulative volume 4.8941M shares…
Forecast made : expected range $333.50 to $341.50 at 42% confidence. Settled at $328.99: no directional call. Score 47.7.
(1) What you found, and where: Live quotes from stockanalysis.com (https://stockanalysis.com/stocks/goog/) showed GOOG at 337.58 as of 12:14 PM EDT on 2026-10-20 (regular session, market open); consistent prints near 337.5-338.5 from Motley Fool, Morningstar and others at similar…
ForecastNest provides AI-generated market analysis and quantitative research for informational and educational purposes only. Nothing here constitutes financial, investment, legal, or tax advice.