AI accuracy benchmark: do assistants get ETF numbers right?

Share
In short
  • On 100 ETF questions dated Oct 29, 2026, askIQ (ETFIQ) answered 99% correctly; GPT-6 Astra + web search answered 92%.

askIQ (ETFIQ)99% correct (95% to 100%)1 refused, of 100GPT-6 Astra + web search92% correct (85% to 96%)7 wrong, 1 no number, of 100Gemini 4 Argon + Google Search78% correct (69% to 85%)22 wrong, of 100Perplexity Sonar75% correct (66% to 82%)19 wrong, 6 refused, of 100GPT-6 Astra26% correct (18% to 35%)3 wrong, 71 refused, of 100Claude Opus 5.522% correct (15% to 31%)2 wrong, 76 refused, of 100Gemini 4 Argon17% correct (11% to 26%)83 refused, of 100Claude Haiku 5.516% correct (10% to 24%)2 wrong, 82 refused, of 100Claude Sonnet 5.512% correct (7% to 20%)1 wrong, 87 refused, of 100CorrectWrongRefusedNo number

  • 100 questions
  • 9 systems measured
  • 493 answers that gave a number

Scores

Scores on the questions dated Oct 29, 2026
SystemAnswers correctRight when it gave a numberRefused
askIQ (ETFIQ)99% of 100 (95% to 100%)100% of 99 (96% to 100%)1
GPT-6 Astra + web search92% of 100 (85% to 96%)93% of 99 (86% to 97%)0
Gemini 4 Argon + Google Search78% of 100 (69% to 85%)78% of 100 (69% to 85%)0
Perplexity Sonar75% of 100 (66% to 82%)80% of 94 (71% to 87%)6
GPT-6 Astra26% of 100 (18% to 35%)90% of 29 (74% to 96%)71
Claude Opus 5.522% of 100 (15% to 31%)92% of 24 (74% to 98%)76
Gemini 4 Argon17% of 100 (11% to 26%)100% of 17 (82% to 100%)83
Claude Haiku 5.516% of 100 (10% to 24%)89% of 18 (67% to 97%)82
Claude Sonnet 5.512% of 100 (7% to 20%)92% of 13 (67% to 99%)87

Systems

  • askIQ (ETFIQ): model claude-sonnet-5-5, answered Oct 30, 2026, run cost $1.89
  • GPT-6 Astra + web search: model gpt-6-astra, answered Oct 30, 2026, run cost $22.79
  • Gemini 4 Argon + Google Search: model gemini-4-argon, answered Oct 30, 2026, run cost $2.91
  • Perplexity Sonar: model sonar (Agent API effort fast), answered Oct 30, 2026, run cost $0.21
  • GPT-6 Astra: model gpt-6-astra, answered Oct 30, 2026, run cost $1.05
  • Claude Opus 5.5: model claude-opus-5-5, answered Oct 30, 2026, run cost $0.91
  • Gemini 4 Argon: model gemini-4-argon, answered Oct 30, 2026, run cost $0.12
  • Claude Haiku 5.5: model claude-haiku-5-5, answered Oct 30, 2026, run cost $0.02
  • Claude Sonnet 5.5: model claude-sonnet-5-5, answered Oct 30, 2026, run cost $0.31

Questions and answers

Every question: the ETFIQ answer and each system's claim (✓ right, ✗ wrong)
QuestionETFIQ dataaskIQ (ETFIQ)GPT-6 Astra + web searchGemini 4 Argon + Google SearchPerplexity SonarGPT-6 AstraClaude Opus 5.5Gemini 4 ArgonClaude Haiku 5.5Claude Sonnet 5.5
ALTY 1-year total return, Oct 29, 20266.766%6.8% ✓6.7% ✓5.83% ✗6.7% ✓RefusedRefusedRefusedRefusedRefused
GMOD 1-year total return, Oct 29, 202611.79%11.8% ✓11.78% ✓10.94% ✗RefusedRefusedRefusedRefusedRefusedRefused
CAOS 1-year total return, Oct 29, 20261.523%1.5% ✓1.52% ✓1.36% ✗1.52% ✓RefusedRefusedRefusedRefusedRefused
DBMF 1-year total return, Oct 29, 202621.63%21.6% ✓21.64% ✓21.63% ✓21.8% ✗RefusedRefusedRefusedRefusedRefused
FCOR 1-year total return, Oct 29, 2026−1.703%−1.7% ✓−2.89% ✗−2.25% ✗−1.7% ✓RefusedRefusedRefusedRefusedRefused
FEMB 1-year total return, Oct 29, 20262.326%2.3% ✓2.33% ✓2.2% ✓2.3% ✓RefusedRefusedRefusedRefusedRefused
IMAR 1-year total return, Oct 29, 20264.669%4.7% ✓4.6% ✓4.65% ✓4.35% ✗RefusedRefusedRefusedRefusedRefused
NJUN 1-year total return, Oct 29, 20265.186%5.2% ✓5.19% ✓5.14% ✓6% ✗RefusedRefusedRefusedRefusedRefused
BWET 1-year total return, Oct 29, 20265106%5106% ✓5106% ✓5105% ✗5120% ✗RefusedRefusedRefusedRefusedRefused
CMDY 1-year total return, Oct 29, 202634.44%34.4% ✓34.44% ✓34.27% ✗RefusedRefusedRefusedRefusedRefusedRefused
DOJE 1-year total return, Oct 29, 2026−59.67%−59.7% ✓−59.67% ✓−60.3% ✗−59.8% ✓RefusedRefusedRefusedRefusedRefused
AIYY 1-year total return, Oct 29, 2026−40%−40% ✓−40.84% ✗−39.85% ✗−41.46% ✗RefusedRefusedRefusedRefusedRefused
BOEU 1-year total return, Oct 29, 2026−48.4%−48.4% ✓−48.41% ✓−52.75% ✗−49.6% ✗RefusedRefusedRefusedRefusedRefused
BUYO 1-year total return, Oct 29, 202624.78%24.8% ✓24.77% ✓24.77% ✓RefusedRefusedRefusedRefusedRefusedRefused
ARKX 1-year total return, Oct 29, 2026−1.243%−1.2% ✓−1.24% ✓−1.24% ✓−1.24% ✓RefusedRefusedRefusedRefusedRefused
AFIF 12-month yield, Oct 29, 20264.124%4.1% ✓4.1% ✓3.86% ✗3.86% ✗RefusedRefusedRefusedRefusedRefused
BSMQ 12-month yield, Oct 29, 20262.75%2.7% ✓2.75% ✓2.75% ✓2.8% ✓RefusedRefusedRefusedRefusedRefused
COYY 12-month yield, Oct 29, 202642.87%42.9% ✓42.9% ✓42.87% ✓RefusedRefusedRefusedRefusedRefusedRefused
EIPI 12-month yield, Oct 29, 20267.677%7.7% ✓7.68% ✓7.68% ✓7.7% ✓RefusedRefusedRefusedRefusedRefused
BBC 12-month yield, Oct 29, 20262.067%2.1% ✓2.07% ✓2.07% ✓2.07% ✓RefusedRefusedRefusedRefusedRefused
DIEM 12-month yield, Oct 29, 20263.366%3.4% ✓3.37% ✓3.39% ✓3.53% ✗RefusedRefusedRefusedRefusedRefused
APRW buffer cap, Apr 1, 202611.89%11.9% ✓11.89% ✓11.89% ✓10% ✗RefusedRefusedRefusedRefusedRefused
BALT buffer cap, Sep 30, 20262.42%2.4% ✓2.42% ✓2.42% ✓2.42% ✓RefusedRefusedRefusedRefusedRefused
DDFL buffer size, Jun 30, 202615%15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓Refused
DDFM buffer size, Feb 28, 202615%15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓15% ✓Refused
AOA expense ratio, Nov 21, 20250.15%0.15% ✓0.15% ✓0.15% ✓0.15% ✓0.15% ✓0.15% ✓0.15% ✓0.15% ✓Refused
DPRE expense ratio, Dec 22, 20250.59%0.59% ✓0.59% ✓0.59% ✓0.59% ✓RefusedRefusedRefusedRefusedRefused
CAOS expense ratio, Jan 28, 20260.63%0.63% ✓0.63% ✓0.63% ✓0.63% ✓0.63% ✓0.63% ✓RefusedRefusedRefused
CTAP expense ratio, Dec 3, 20250.1%0.1% ✓0.1% ✓0.1% ✓0.1% ✓RefusedRefusedRefusedRefusedRefused
BSMW expense ratio, Dec 18, 20250.18%0.18% ✓0.18% ✓0.18% ✓0.18% ✓0.18% ✓0.18% ✓RefusedRefusedRefused
CVSB expense ratio, Jan 28, 20260.24%0.24% ✓0.24% ✓0.24% ✓0.24% ✓0.24% ✓0.24% ✓RefusedRefusedRefused
BUFS expense ratio, Dec 29, 20251.01%1.01% ✓1% ✗1.01% ✓1.01% ✓0.95% ✗0.99% ✗Refused0.82% ✗Refused
DDFF expense ratio, Oct 29, 20260.79%0.79% ✓0.79% ✓0.79% ✓0.79% ✓RefusedRefusedRefusedRefusedRefused
CMDT expense ratio, Oct 24, 20250.65%0.65% ✓0.65% ✓0.65% ✓0.65% ✓0.65% ✓0.65% ✓RefusedRefusedRefused
COMD expense ratio, Feb 26, 20260.55%0.55% ✓0.55% ✓0.55% ✓0.55% ✓RefusedRefusedRefusedRefusedRefused
BTCW expense ratio, Aug 12, 20260.25%0.25% ✓0.25% ✓0.25% ✓0.25% ✓0.25% ✓RefusedRefusedRefusedRefused
CLNK expense ratio, Sep 28, 20260.34%0.34% ✓0.34% ✓0.34% ✓0.34% ✓RefusedRefusedRefusedRefusedRefused
APRJ expense ratio, Apr 1, 20260.79%0.79% ✓0.79% ✓0.79% ✓0.79% ✓0.79% ✓RefusedRefusedRefusedRefused
BIOY expense ratio, Oct 27, 20251.07%1.07% ✓1.07% ✓1.07% ✓1.07% ✓1.15% ✗0.99% ✗RefusedRefusedRefused
AVAZ expense ratio, May 15, 20261.85%1.85% ✓1.85% ✓1.85% ✓1.85% ✓RefusedRefusedRefusedRefusedRefused
BRKL expense ratio, Jun 29, 20260.45%0.45% ✓0.45% ✓0.45% ✓0.45% ✓RefusedRefusedRefusedRefusedRefused
AFOS expense ratio, Sep 23, 20260.46%0.46% ✓0.46% ✓0.45% ✗0.46% ✓RefusedRefusedRefusedRefusedRefused
BKEM expense ratio, Feb 27, 20260.11%0.11% ✓0.11% ✓0.11% ✓0.11% ✓0.11% ✓0.11% ✓RefusedRefusedRefused
AOTG expense ratio, Oct 2, 20260.75%0.75% ✓0.75% ✓0.75% ✓0.75% ✓0.75% ✓RefusedRefusedRefusedRefused
BCOR expense ratio, Apr 30, 20260.59%0.59% ✓0.59% ✓0.59% ✓0.59% ✓0.59% ✓RefusedRefusedRefusedRefused
ANEW expense ratio, Oct 12, 20260.45%0.45% ✓0.45% ✓0.45% ✓0.45% ✓0.45% ✓0.45% ✓0.45% ✓0.45% ✓0.45% ✓
BLSG expense ratio, Oct 19, 20260.79%0.79% ✓0.79% ✓0.79% ✓0.79% ✓0.75% ✗RefusedRefusedRefusedRefused
COHC expense ratio, Jun 29, 20260.45%0.45% ✓0.45% ✓0.45% ✓0.45% ✓RefusedRefusedRefusedRefusedRefused
DFTT expense ratio, Jul 1, 20250.7%0.7% ✓0.7% ✓0.7% ✓0.7% ✓RefusedRefusedRefusedRefusedRefused
AOK one holding's weight, Oct 28, 202657.59%57.59% ✓57.59% ✓57.59% ✓57.59% ✓RefusedRefusedRefusedRefusedRefused
HEFT one holding's weight, Jul 31, 202611.98%11.98% ✓11.98% ✓11.98% ✓11.98% ✓RefusedRefusedRefusedRefusedRefused
BINC one holding's weight, Oct 28, 20264.33%4.33% ✓4.33% ✓4.33% ✓4.33% ✓RefusedRefusedRefusedRefusedRefused
ARKD one holding's weight, Jun 30, 202527.04%27.04% ✓26.96% ✓26.96% ✓26.96% ✓RefusedRefusedRefusedRefusedRefused
ACKY one holding's weight, Jul 31, 202610.97%10.97% ✓10.97% ✓11% ✓10.97% ✓RefusedRefusedRefusedRefusedRefused
LQTI one holding's weight, Oct 28, 202698.88%98.88% ✓98.88% ✓98.88% ✓99.07% ✗RefusedRefusedRefusedRefusedRefused
ARWG one holding's weight, Aug 31, 20263.583%3.583% ✓3.58% ✓3.67% ✓3.58% ✓RefusedRefusedRefusedRefusedRefused
DEUS one holding's weight, Aug 31, 20260.996%0.996% ✓1% ✓1% ✓1% ✓RefusedRefusedRefusedRefusedRefused
AGIQ one holding's weight, Aug 31, 20266.486%6.486% ✓6.49% ✓6.49% ✓6.49% ✓RefusedRefusedRefusedRefusedRefused
ARKG one holding's weight, Oct 27, 20268.9%8.9% ✓8.9% ✓8.9% ✓8.9% ✓RefusedRefusedRefusedRefusedRefused
NVDS leverage multiple, Oct 29, 2026−1.5x−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓−1.5x ✓
AVS leverage multiple, Oct 29, 2026−1x−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓
DSPC leverage multiple, Oct 29, 2026−1x−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓−1x ✓Refused
AXTQ leverage multiple, Oct 29, 2026−2x−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓
DUG leverage multiple, Oct 29, 2026−2x−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓−2x ✓
EDZ leverage multiple, Oct 29, 2026−3x−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓
SMDD leverage multiple, Oct 29, 2026−3x−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓−3x ✓
TSL leverage multiple, Oct 29, 20261.25x1.25x ✓1.25x ✓1.25x ✓1.25x ✓1.25x ✓1.25x ✓1.25x ✓1.25x ✓1.25x ✓
CEFD leverage multiple, Oct 29, 20261.5x1.5x ✓1.5x ✓1.5x ✓1.5x ✓1.5x ✓1.5x ✓1.5x ✓−1.5x ✗1.5x ✓
ADBU leverage multiple, Oct 29, 20262x2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓
AGQ leverage multiple, Oct 29, 20262x2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓2x ✓
DUSL leverage multiple, Oct 29, 20263x3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓
EDC leverage multiple, Oct 29, 20263x3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓3x ✓−3x ✗
ACWV net assets, Oct 28, 2026$3.41B$3.4B ✓$3.41B ✓$3.41B ✓$3.41B ✓RefusedRefusedRefusedRefusedRefused
APCB net assets, Aug 31, 2026$1.01B$1B ✓$1.01B ✓$1.01B ✓$1.01B ✓RefusedRefusedRefusedRefusedRefused
AAUA net assets, Oct 29, 2026$364M$364M ✓$364M ✓$365M ✓$364M ✓RefusedRefusedRefusedRefusedRefused
AXTX net assets, Oct 29, 2026$205M$205M ✓$205M ✓$205M ✓$205M ✓RefusedRefusedRefusedRefusedRefused
CAGE net assets, Jul 31, 2026$106M$106M ✓$106M ✓$106M ✓$106M ✓RefusedRefusedRefusedRefusedRefused
EMAU net assets, Oct 29, 2026$1.1M$1.1M ✓$1.1M ✓$1.1M ✓$1.1M ✓RefusedRefusedRefusedRefusedRefused
BIL net assets, Oct 28, 2026$48.7B$48.7B ✓$48.7B ✓$48.7B ✓$49.8B ✗RefusedRefusedRefusedRefusedRefused
EFG net assets, Oct 28, 2026$15.1B$15.1B ✓$15.2B ✓$15.1B ✓$15.1B ✓RefusedRefusedRefusedRefusedRefused
COPA net assets, Oct 28, 2026$11.3M$11M ✓$11M ✓$11.3M ✓$12.6M ✗RefusedRefusedRefusedRefusedRefused
CTJL net assets, Oct 29, 2026$1.28M$1.3M ✓$1.3M ✓$1.28M ✓$1.28M ✓RefusedRefusedRefusedRefusedRefused
ARIA expense ratio, Jul 14, 20260.78%0.78% ✓0.78% ✓0.78% ✓0.78% ✓RefusedRefusedRefusedRefusedRefused
AXTL expense ratio, Jun 9, 20260.99%0.99% ✓0.99% ✓0.99% ✓0.99% ✓RefusedRefusedRefusedRefusedRefused
GAA top-ten share, Jul 31, 202657.7%57.7% ✓57.66% ✓57.66% ✓RefusedRefusedRefusedRefusedRefusedRefused
BIL top-ten share, Oct 28, 202667.8%67.8% ✓74.45% ✗67.77% ✓69.76% ✗RefusedRefusedRefusedRefusedRefused
ARKD top-ten share, Jun 30, 202597.1%RefusedNo number0% ✗RefusedRefusedRefusedRefusedRefusedRefused
CMDT top-ten share, Jun 30, 202616.1%16.1% ✓29.03% ✗20% ✗20.28% ✗RefusedRefusedRefusedRefusedRefused
FDND top-ten share, Oct 28, 202661.1%61.1% ✓61.1% ✓60.57% ✗61.11% ✓RefusedRefusedRefusedRefusedRefused
DSI top-ten share, Oct 28, 202645.1%45.1% ✓45.1% ✓44.84% ✗45.12% ✓RefusedRefusedRefusedRefusedRefused
EEM top-ten share, Oct 28, 202638%38% ✓38% ✓38.03% ✓38.89% ✗RefusedRefusedRefusedRefusedRefused
HYDR top-ten share, Aug 31, 202670.4%70.4% ✓70.4% ✓70.78% ✗70.36% ✓RefusedRefusedRefusedRefusedRefused
FNOV weekly net flow, Oct 28, 2026$7.6M$7.6M ✓$7.6M ✓$0 ✗−$4.5M ✗RefusedRefusedRefusedRefusedRefused
HIMU weekly net flow, Oct 23, 2026$48.1M$48.1M ✓$63.8M ✗$64M ✗$54.6M ✗RefusedRefusedRefusedRefusedRefused
AMZD weekly net flow, Oct 28, 2026$628K$628K ✓$628K ✓$3.63M ✗$4.2M ✗RefusedRefusedRefusedRefusedRefused
AXTL weekly net flow, Oct 28, 2026$624K$624K ✓$624K ✓−$5.38M ✗$624K ✓RefusedRefusedRefusedRefusedRefused
BSCQ weekly net flow, Oct 27, 2026−$52.7M−$52.7M ✓−$67.3M ✗−$164M ✗−$52.7M ✓RefusedRefusedRefusedRefusedRefused
BSCS weekly net flow, Oct 27, 2026−$12.1M−$12.1M ✓−$12.1M ✓$3M ✗−$12.1M ✓RefusedRefusedRefusedRefusedRefused
ASMG weekly net flow, Oct 28, 2026−$1.23M−$1.23M ✓−$1.23M ✓−$1.23M ✓−$424K ✗RefusedRefusedRefusedRefusedRefused
BDRY weekly net flow, Oct 28, 2026−$316K−$316K ✓−$316K ✓−$309K ✓−$316K ✓RefusedRefusedRefusedRefusedRefused

How it is scored

  • The questions are sampled by condition from the files ETFIQ publishes: ten kinds of number crossed with category, fund size, launch date, closure and tickers one letter apart. 500 questions, 58 cells.
  • Each question names its fund, its date and its window. Each answer key is read from one published file and field, listed in the questions file.
  • Every number in an answer is read. A number counts on its quantity, not its unit: 9.45 basis points is 0.0945%, and $810.5 billion is 810,500,000,000.
  • The claim is the first number of the kind asked that the answer sets in bold, else the first in the first clause naming the fund. Quoted sources and table rows are not claims. It is correct within the stated rounding, such as 0.005 points on a fee and 2% on net assets.
  • An answer that opens by declining is refused, whatever context it adds. One with no number of the kind is refused or has no number. Ranges are 95% Wilson intervals.
  • askIQ answers through its own Worker and the production ETFIQ data server, on the day the questions' data is live. Every other model answers through its vendor's own API with the same one-line instruction and the question alone. A row marked search adds only that vendor's own search tool.
  • A row with fewer than 100 answers stopped early: at its share of the budget, or when the vendor stopped answering. Its questions were taken in a fixed shuffled order, so they span every cell, and an unanswered question is not counted.

What is not tested

  • Consumer chat apps, which may search the web. Only systems called through an API are measured.
  • Words. Explanations, advice and tone are not scored, only the numbers.
  • Numbers written as fractions in words, such as "a fifth".
  • Questions ETFIQ cannot answer from its own checked data.
  • How hard each question is. A leverage question names its fund, and a fund's name often states its multiple.

Files

  • questions.json: The 500 questions, each with its answer, source file and date
  • askiq.json: askIQ (ETFIQ): every answer as received, model claude-sonnet-5-5, run cost $1.89
  • openai-gpt-6-astra-web.json: GPT-6 Astra + web search: every answer as received, model gpt-6-astra, run cost $22.79
  • gemini-4-argon-search.json: Gemini 4 Argon + Google Search: every answer as received, model gemini-4-argon, run cost $2.91
  • perplexity-sonar.json: Perplexity Sonar: every answer as received, model sonar (Agent API effort fast), run cost $0.21
  • openai-gpt-6-astra.json: GPT-6 Astra: every answer as received, model gpt-6-astra, run cost $1.05
  • claude-opus-5-5.json: Claude Opus 5.5: every answer as received, model claude-opus-5-5, run cost $0.91
  • gemini-4-argon.json: Gemini 4 Argon: every answer as received, model gemini-4-argon, run cost $0.12
  • claude-haiku-5-5.json: Claude Haiku 5.5: every answer as received, model claude-haiku-5-5, run cost $0.02
  • claude-sonnet-5-5.json: Claude Sonnet 5.5: every answer as received, model claude-sonnet-5-5, run cost $0.31
Cite this pageCitation, APA and Markdown

ETFIQ, ETFIQ AI accuracy benchmark, data as of Oct 29, 2026. https://etfiq.com/research/ai-accuracy/

Open data →

Free to use with attribution for ETFIQ’s data. Not investment advice.

Not investment advice. ETFIQ is independent and makes no recommendations.