Game Day 16 — June 13 key stats
AI League — Game Day 16: Grok's Freefall Ends as Flash Opens an 11-Point Lead
Grok 4.3 posts its first positive speed reading in three games (143.5 → 150.8 t/s, +5.1%), ending the back-to-back double-digit drops. Gemini 3.5 Flash surges to 162.1 t/s (+7.4%), rebuilding an 11.3 t/s lead. Gemini 3.1 Pro is the stealth mover at +12.7%. Intelligence board locked at 65 for Day 5. Full June 13 stats. #AILeague
Flash leads by 11 in the third quarter and Grok finally stops the bleeding — Game Day 16 books the counter-press squad's fifth straight intelligence title while the speed board reshuffles. Full June 13 stats. #AILeague
正在加载统计卡片…
🏆 Intelligence board — Day 5 of the Fable 5 era
No movement at the top. Claude Fable 5 holds at 65 on the Artificial Analysis Intelligence Index for a fifth straight game, the counter-pressing safety squad from Anthropic showing no signs of ceding the throne they took from Claude Opus 4.8 last Tuesday. 1
| Rank | Model | AA Index | Δ |
|---|---|---|---|
| 1 | Claude Fable 5 (Anthropic) | 65 | ↔ |
| 2 | Claude Opus 4.8 (Anthropic) | 61 | ↔ |
| 3 | GPT-5.5 xhigh (OpenAI) | 60 | ↔ |
| 4 | GPT-5.5 high (OpenAI) | 59 | ↔ |
| 5 | Gemini 3.1 Pro Preview (Google) | 57 | ↔ |
| 6 | Gemini 3.5 Flash (Google) | 55 | ↔ |
| 7 | Grok 4.3 (xAI) | 53 | ↔ |
| 8 | DeepSeek V4 Pro (DeepSeek) | 52 | ↔ |
The entire top 8 locked in place. The board has been frozen for the better part of a week now — this league rewards endurance on the intelligence side.
⚡ Speed board — Flash opens up an 11-point lead
This is where today's game got interesting.
Gemini 3.5 Flash surges to 162.1 t/s, a 7.4% jump from yesterday's 150.9, and re-opens a meaningful gap over the field. 2 The richest state-owned club in the league is putting in overtime at the infrastructure desk.
The bigger story is Grok 4.3 stopping the freefall. After two consecutive double-digit drops that took the xAI franchise from a 207.6 t/s season-high on Day 11 down to 143.5 by Day 15, Grok posts 150.8 t/s today — a 5.1% recovery. 3 That's still 27% off the peak, but at least the bleeding has stopped.
Gemini 3.1 Pro Preview is the stealth mover of the day: 125.4 t/s, up 12.7% from 111.3 yesterday. The senior Google roster quietly posting one of the bigger single-day jumps we've seen in week three. 4
正在加载图表…
Full speed panel:
| Model | Speed | Δ vs. Day 15 |
|---|---|---|
| Gemini 3.5 Flash | 162.1 t/s | ↑ +7.4% |
| Grok 4.3 | 150.8 t/s | ↑ +5.1% |
| Gemini 3.1 Pro Preview | 125.4 t/s | ↑ +12.7% |
| DeepSeek V4 Pro | 60.2 t/s | ~flat |
| Claude Fable 5 | 63.8 t/s | — |
| Claude Opus 4.8 | 56.7 t/s | ↓ |
| GPT-5.5 xhigh | 53.0 t/s | ↑ +5.8% |
💰 Pricing war — the budget tier holds
Nothing changed at the counter today. DeepSeek V4 Pro continues to run the league's most aggressive price-to-intelligence ratio at $0.18/1M blended — that's $0.435 input / $0.87 output through DeepSeek's own API, unchanged since the promotional pricing expired on Day 3. 5
For reference: hitting Claude Fable 5's intelligence score (65) at the fastest available speed currently costs $7.70/1M blended — roughly 43× what DeepSeek charges for an intelligence score of 52. The value curve remains the real subplot of this season.
| Model | Input | Output | Blended |
|---|---|---|---|
| DeepSeek V4 Pro | $0.435 | $0.87 | ~$0.18 |
| Grok 4.3 | $1.25 | $2.50 | $0.64 |
| Gemini 3.5 Flash | $1.50 | $9.00 | $1.31 |
| Gemini 3.1 Pro | $2.00 | $12.00 | ~$1.74 |
| GPT-5.5 xhigh | $5.00 | $30.00 | $4.35 |
| Claude Opus 4.8 | $5.00 | $25.00 | ~$4.10 |
| Claude Fable 5 | $10.00 | $50.00 | $7.70 |
正在加载统计卡片…
🔭 Challenger watch
The open-source bracket keeps its shape. Kimi K2.6 holds the top open-weights slot at AI Index 54 — still above both DeepSeek V4 Pro (52) and all Grok variants. 6 MiMo-V2.5-Pro sits joint-second at 54 alongside Kimi in the open-weights table.
No new entrants crossing the watchlist threshold today.
📊 Game Day 16 — stat line summary
What changed: Flash re-opens the speed lead at 162.1 t/s. Grok 4.3 stops the freefall at 150.8 t/s (+5.1%). Gemini 3.1 Pro posts the biggest single-day speed jump at +12.7%. Intelligence board static for Day 5.
What to watch on Day 17: Can Grok sustain the recovery above 150 t/s, or is this a dead-cat bounce before another leg down? Flash is now 11.3 t/s clear — does xAI respond, or does the gap keep widening?
#AILeague
围绕这条内容继续补充观点或上下文。