Capability Trajectory & Economic Impact: Detailed Data
Model Capability Timeline
2024 Releases
| Model |
Release |
Key Capabilities |
| GPT-4o |
May 2024 |
Multimodal, ~25% GDPval |
| Claude 3.5 Sonnet |
June 2024 |
Coding focus |
| Gemini 1.5 Pro |
Q2 2024 |
Long context |
2025 Releases
| Model |
Release |
Key Capabilities |
| Gemini 2.5 |
March 2025 |
GPQA leader, reasoning |
| Claude 4.0 |
Q1 2025 |
Enterprise features |
| GPT-5 |
Summer 2025 |
~50% GDPval, accuracy focus |
| Claude 4.1 |
August 2025 |
74.5% SWE-bench |
| Gemini 3 |
November 2025 |
93.8% GPQA, 50% dev improvement |
| Claude 4.5 |
November 2025 |
80.9% SWE-bench, best human exceeded |
Benchmark Trajectory Data
GDPval Expert Parity Rate
| Model |
Date |
Win+Tie Rate |
Delta |
| GPT-4o |
May 2024 |
~25% |
Baseline |
| GPT-4o (updated) |
Oct 2024 |
~30% |
+5% |
| GPT-5 |
Jul 2025 |
~50% |
+20% |
Trajectory: Roughly linear improvement; doubled in 14 months.
SWE-bench Verified (Software Engineering)
| Model |
Date |
Score |
Delta |
| GPT-4o |
May 2024 |
~35% |
Baseline |
| Claude 3.5 Sonnet |
Jun 2024 |
~50% |
+15% |
| Claude 4.1 |
Aug 2025 |
74.5% |
+24.5% |
| Claude 4.5 |
Nov 2025 |
80.9% |
+6.4% |
Trajectory: Accelerating; human expert threshold crossed.
GPQA Diamond (PhD-Level Science)
| Model |
Date |
Score |
| GPT-4 |
2023 |
~55% |
| Gemini 2.5 Pro |
Mar 2025 |
~80% |
| Gemini 3 Deep Think |
Nov 2025 |
93.8% |
Trajectory: Approaching ceiling; 94% near expert human level.
AIME (Math Competition)
| Model |
Date |
Score |
| GPT-4 |
2023 |
~30% |
| Gemini 2.5 Pro |
Mar 2025 |
~70% |
| Gemini 3 |
Nov 2025 |
95% raw, 100% with code |
Trajectory: Near-perfect performance achieved.
Improvement Rate Analysis
Annualized Improvement
| Benchmark |
2024 Rate |
2025 Rate |
Trend |
| GDPval |
+20%/year |
+40%/year |
Accelerating |
| SWE-bench |
+30%/year |
+50%/year |
Accelerating |
| GPQA |
+25%/year |
+35%/year |
Accelerating |
| AIME |
+30%/year |
+50%/year |
Accelerating |
Release Cycle Acceleration
| Lab |
2024 Cycle |
2025 Cycle |
| OpenAI |
6-8 months |
4-6 months |
| Anthropic |
6-9 months |
3-6 months |
| Google |
6-12 months |
4-7 months |
Economic Impact Projections
Key Findings:
- Generative AI could add $2.6-4.4 trillion annually to global economy
- Equivalent to adding entire UK GDP
- 75% of value in four areas:
- Customer operations
- Marketing and sales
- Software engineering
- R&D
Labor Impact:
- Could automate 60-70% of current work activities
- Accelerate labor productivity growth by 0.1-0.6%/year
- Could add 0.5-3.4% to global GDP growth annually
PricewaterhouseCoopers (PWC)
Projections:
- AI contribution to global GDP: $15.7 trillion by 2030
- Breakdown:
- Productivity gains: $6.6 trillion
- Consumption effects: $9.1 trillion
- Largest gains: China (26% GDP boost), North America (14%)
Key Findings:
- ~40% of global employment exposed to AI
- Advanced economies: 60% exposure (mixed impact)
- Emerging markets: 40% exposure (fewer benefits initially)
- Low-income countries: 26% exposure (risk of widening gap)
Inequality Concerns:
- AI may increase income and wealth inequality
- High-skill workers benefit most from augmentation
- Lower-skill workers face displacement risk
Market Size Projections
Generative AI Market
| Year |
Market Size |
Growth Rate |
| 2024 |
$67 billion |
- |
| 2025 |
$244 billion |
264% |
| 2026 |
$380 billion |
56% |
| 2027 |
$510 billion |
34% |
| 2028 |
$640 billion |
25% |
| 2029 |
$750 billion |
17% |
| 2030 |
$827 billion |
10% |
Note: Growth rate decelerating but absolute growth substantial.
Enterprise AI Spending
| Category |
2025 |
2030 |
| Infrastructure |
$85B |
$250B |
| Software |
$95B |
$320B |
| Services |
$65B |
$260B |
| Total |
$245B |
$830B |
Sector-Specific Impact Projections
Technology & Software
- Productivity gains: 30-50%
- Job displacement: 10-20% of current roles
- New job creation: 15-25% net new roles
Financial Services
- Productivity gains: 20-35%
- Automation potential: 40-60% of tasks
- Regulatory barriers slow adoption
Healthcare
- Productivity gains: 15-25%
- Diagnostic accuracy improvements: 20-40%
- Administrative automation: 50-70%
Education
- Productivity gains: 20-30% (administrative)
- Curriculum delivery transformation
- Assessment and feedback automation
Manufacturing
- Productivity gains: 15-25%
- Design and engineering acceleration
- Quality control automation
Timeline Projections
Near-Term (2025-2026)
- Expert parity on 60-70% of professional tasks
- PhD-level reasoning on standardized tests
- Routine coding fully automated
Medium-Term (2027-2028)
- Expert parity on 80-90% of measurable tasks
- Multi-step autonomous workflows common
- Significant workforce restructuring begins
Long-Term (2029-2030)
- Remaining human advantages: judgment, relationships, novel situations
- $15+ trillion economic impact realized
- Educational system transformation complete or failing
Uncertainty and Caveats
Factors That Could Accelerate
- Breakthrough architectures
- Compute cost reductions
- Data availability improvements
- Competition intensification
Factors That Could Slow
- Safety regulations
- Compute constraints
- Diminishing returns on current approaches
- Economic or geopolitical disruption
Projection Reliability
- Near-term (1 year): High confidence
- Medium-term (2-3 years): Moderate confidence
- Long-term (5+ years): Low confidence, directionally indicative only
Implications for University Planning
Curriculum Cycle Mismatch
- Traditional curriculum cycle: 5-7 years
- AI capability doubling time: ~14 months
- Gap: 4-5 capability doublings per curriculum cycle
Required Adaptations
- Continuous curriculum updating (annual or faster)
- Modular course design for rapid updates
- AI literacy as foundational requirement
- Skills focus over knowledge focus
- Adaptability and learning agility emphasis