Core Conclusion
xAI released a heavyweight model in the most xAI way possible: no press conference, no blog post, just dropped it directly in the API.
Grok 4.3 has quietly gone live on platforms like Venice, supporting 1 million token context, function calling, multimodal input, and native X search. It achieved a score of 53 on the Artificial Analysis Intelligence Index, surpassing Muse Spark, Claude Sonnet 4.6, and previous Grok iterations. API pricing was adjusted simultaneously: input dropped from $2.10 to $1.25/M tokens (40% cut), output cut by 60%.
Benchmark Performance
Artificial Analysis Intelligence Index
| Model | AA Index | Notes |
|---|---|---|
| GPT-5.5 Pro | ~60+ | Current leader |
| Grok 4.3 | 53 | Surpassed Muse Spark, Sonnet 4.6 |
| Muse Spark | <53 | Surpassed by Grok 4.3 |
| Claude Sonnet 4.6 | <53 | Surpassed by Grok 4.3 |
| Gemini 3.1 Pro | ~50 | Close to Grok 4.3 |
Vals Index Rankings
| Benchmark | Grok 4.3 Rank | Notes |
|---|---|---|
| Overall | #13 | Above average |
| CaseLaw | #1 | Top-tier legal reasoning |
| CorpFin | #1 | Top-tier corporate finance analysis |
| General Coding | Weak | Not a strength |
GDPval-AA Benchmark
Grok 4.3’s most significant improvement is in real-world Agent tasks — on the GDPval-AA benchmark, Grok 4.3’s agentic capability score increased substantially. This is the core metric for measuring “can AI complete tasks independently.”
Pricing Strategy Analysis
| Item | Grok 4.3 | Change |
|---|---|---|
| Input Price | $1.25/M tokens | ↓ 40% |
| Output Price | Significantly reduced | ↓ 60% |
| Context Window | 1M tokens | Same as previous |
This pricing strategy is extremely aggressive. The $1.25/M token input price is already lower than most mid-tier models, yet Grok 4.3’s performance sits in the top tier. xAI is clearly pursuing a “cost-performance route” — delivering near Claude Opus 4.7 performance at prices approaching DeepSeek V4.
Horizontal Comparison with Competitors
| Dimension | Grok 4.3 | Claude Sonnet 4.6 | GPT-5.5 | DeepSeek V4 |
|---|---|---|---|---|
| AA Index | 53 | <53 | ~60+ | N/A |
| Input Price | $1.25/M | ~$3/M | ~$5/M | ~$0.15/M |
| Legal Reasoning | #1 | Strong | Strong | Medium |
| Financial Analysis | #1 | Strong | Strong | Medium |
| General Coding | Weak | Strong | Strong | Strong |
| Agent Capability | Significantly improved | Strong | Strong | Strong |
Landscape Assessment
Grok 4.3’s release signals several things:
- xAI is transitioning from “chaser” to “cost-performance leader”: An AA index of 53 with $1.25 pricing delivers far better value than Claude and GPT
- Clear advantage in specialized domains: #1 rankings in CaseLaw and CorpFin indicate Grok 4.3 has unique advantages in legal and finance verticals
- Silent launch shows xAI prioritizes product over marketing: This is both a strength (pragmatic) and weakness (low visibility)
How to Use This
- Legal/Finance professionals: Grok 4.3’s #1 rankings in CaseLaw and CorpFin are worth attention — potentially the most cost-effective specialized model choice
- API users: $1.25/M input pricing + 53-point performance makes this the cheapest option among first-tier models
- Agent developers: The substantial improvement on GDPval-AA means Grok 4.3’s reliability in Agent scenarios has increased significantly — worth testing