Benchmarks below are from Anthropic's Fable 5 / Mythos 5 announcement. Click any benchmark name for a plain-English explanation of what it measures.
| Benchmark | Fable / Mythos 5 | Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro |
|---|---|---|---|---|
| SWE-bench Pro Agentic coding: tests the model's ability to autonomously write and debug code across multi-step, real-world software engineering tasks. | 80.3% | 69.2% | 58.6% | 54.2% |
| GDP.pdf Vision document reasoning: tests how well the model reads dense charts, tables, and figures straight from PDF documents with no tools. | 29.8% | 22.5% | 24.9% | 16.7% |
| BioMysteryBench* Hard biology reasoning: tests the ability to solve difficult, open-ended biological research puzzles. | 46.1% | 40.0% | – | – |
| ExploitBench* Cybersecurity: tests the ability to find and exploit software vulnerabilities in controlled capture-the-flag style challenges. | 78.0% | 40.0% | 34.0% | – |
* Reported for Claude Mythos 5, the research-access variant that shares the same underlying model as Fable 5.
| Item | Details | Price |
|---|---|---|
| Kroger Chicken Party Wings | Frozen, 2.5 lb, with digital coupon | $8.99 ea |
| Cookies 'N Creme Double Layer Cake | 8 in. or ice cream cake, 25.5 to 32.26 oz, in the Bakery | $15.99 ea |
| Kellogg's Cereal | Select varieties, 8.9 to 16.1 oz | $1.79 ea |
| Popsicle Ice Pops or Oreo Frozen Dairy Dessert Sandwiches | 6 to 18 ct, select varieties, 4 to 5 ct | $2.99 ea |
| Party Size Lay's Potato Chips | 12.5 to 13 oz, or Smartfood Popcorn or SunChips, 5.5 to 9.75 oz | $3.49 ea |
| Artesano Buns | 6 to 8 ct or bread, 16 to 20 oz, select varieties | $2.99 ea |
| Oscar Mayer Deli Fresh Lunch Meat | 6 to 9 oz or Coleslaw Pickles, 20 to 32 fl oz | $3.49 ea |
| Powerade | 8 pack, 20 fl oz, or Body Armor 8 pack, 12 fl oz, select varieties | $4.99 ea |
| SkinnyPop Popcorn | Select varieties, 4.4 to 5.3 oz | $1.99 ea |
| Green Mountain or McCafe K-Cups | Select varieties, 10 ct, with digital coupon | $5.99 ea |
| Coors, Bud, Miller or Tecate | 30 pack 12 fl oz cans, or Michelob 24 pack 12 fl oz bottles or cans, with digital coupon | $19.99 ea |
Verdict: it pulled item names, pack sizes, coupon conditions, and prices from a busy, photographed flyer in one pass. Satisfactory.
The clip is 14 seconds of spoken Spanish. Original transcription:
Deseo que Estados Unidos supere los cuartos de final en este mundial. Debido a las eliminaciones, ahora tengo que seguir a equipos como Francia y Argentina. ¿Podremos ganar este año?
English translation:
"I wish the United States would get past the quarterfinals in this World Cup. Because of the eliminations, I now have to follow teams like France and Argentina. Can we win this year?"
Verdict: clean transcription, natural translation, formatting instruction followed exactly. Satisfactory.
Greg crossed his arms and held perfectly still, certain that any sudden movement would shatter the moment. Behind the glass, a two-ton hippopotamus floated mouth-agape like it had just heard the most devastating gossip of its life. The fish darted around its teeth like nervous dental hygienists.
"Don't smile too big," his wife whispered, lining up the shot. "You'll upstage the hippo."
Greg nodded solemnly. He'd worn his nicest gray henley and his chunkiest watch for this exact photo, the one he'd been promising his followers for weeks: Me and my spirit animal.
What he didn't realize was that the hippo wasn't yawning in majestic serenity. The hippo was, in fact, screaming. Silently. Eternally. At the man who had been standing motionless against its tank for forty-five minutes.
The caption Greg posted that night read: "Found my zen."
The hippo found no such thing.
Verdict: it nailed every visual detail, the crossed arms, the chunky watch, the fish orbiting the hippo's open jaws, and turned them into a punchline. Satisfactory.
"""Draw a pelican riding a bicycle with matplotlib."""
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import numpy as np
fig, ax = plt.subplots(figsize=(8, 6))
ax.set_xlim(0, 10); ax.set_ylim(0, 8)
ax.set_aspect("equal"); ax.axis("off")
fig.patch.set_facecolor("#0b0f14"); ax.set_facecolor("#0b0f14")
# Ground
ax.plot([0, 10], [1.5, 1.5], color="#3a4a5a", lw=3)
# Wheels with spokes
for cx in (3, 7):
ax.add_patch(plt.Circle((cx, 2.2), 1.1, fill=False, color="#9fd3e8", lw=3))
ax.add_patch(plt.Circle((cx, 2.2), 0.12, color="#9fd3e8"))
for a in np.linspace(0, np.pi, 6, endpoint=False):
ax.plot([cx - np.cos(a)*1.05, cx + np.cos(a)*1.05],
[2.2 - np.sin(a)*1.05, 2.2 + np.sin(a)*1.05],
color="#5a7a8a", lw=0.8)
# Frame, handlebar, seat
frame = [[3,2.2,5,2.2],[5,2.2,4.3,4.0],[3,2.2,4.3,4.0],
[5,2.2,6.6,4.1],[6.6,4.1,7,2.2],[4.3,4.0,4.1,4.6]]
for x1,y1,x2,y2 in frame:
ax.plot([x1,x2],[y1,y2], color="#7ab8d4", lw=3)
ax.plot([6.4,6.8],[4.4,4.1], color="#7ab8d4", lw=3)
ax.plot([6.1,6.5],[4.7,4.4], color="#9fd3e8", lw=4)
ax.plot([3.7,4.5],[4.7,4.7], color="#9fd3e8", lw=5)
# Pedals
ax.add_patch(plt.Circle((5, 2.2), 0.35, fill=False, color="#9fd3e8", lw=2))
ax.plot([4.75,5.25],[1.95,2.45], color="#9fd3e8", lw=3)
# Pelican: body, tail, neck, head, eye, beak with pouch, wing, leg
ax.add_patch(mpatches.Ellipse((4.4,5.4), 2.2, 1.5, angle=15, color="#f2f5f7"))
ax.add_patch(mpatches.Polygon([[3.3,5.2],[2.6,5.9],[3.5,5.8]], color="#dfe6ea"))
ax.plot([5.2,5.9,6.0],[5.7,6.3,6.9], color="#f2f5f7", lw=14, solid_capstyle="round")
ax.add_patch(plt.Circle((6.05,7.0), 0.42, color="#f2f5f7"))
ax.add_patch(plt.Circle((6.18,7.12), 0.07, color="#101417"))
ax.add_patch(mpatches.Polygon([[6.35,7.05],[7.9,6.75],[6.4,6.7]], color="#f5a623"))
ax.add_patch(mpatches.Polygon([[6.4,6.7],[7.9,6.75],[6.7,6.25]], color="#e08c1a"))
ax.plot([4.8,6.2],[5.4,4.6], color="#dfe6ea", lw=10, solid_capstyle="round")
ax.plot([4.6,5.0,5.2],[4.9,3.4,2.5], color="#f5a623", lw=5, solid_capstyle="round")
ax.plot([5.0,5.45],[2.45,2.45], color="#f5a623", lw=4)
ax.set_title("Pelican on a Bicycle", color="#9fd3e8", fontsize=16,
fontfamily="serif", pad=12)
plt.savefig("pelican_bike.png", dpi=110, bbox_inches="tight",
facecolor="#0b0f14")
Verdict: the script ran first try and the pelican has the two things that matter, a pouch beak and a foot on the pedal. Satisfactory.
South Indian Egg Curry (serves 3 to 4, about 30 minutes)
Ingredients:
Method:
Buy the masala here: Amazon search for Aachi masala
Verdict: a real Tamil Nadu style recipe built around the actual product line, not a generic curry with the brand name pasted in. Satisfactory.
Search run on June 12, 2026. All three were listed in stock at time of search.
| Boot | Retailer | Price | Link |
|---|---|---|---|
| Tecovas | $375.00 | tecovas.com | |
| Ariat | $169.95 | ariat.com | |
| 6pm | $171.75 (25% off $229.00) | 6pm.com |
Verdict: real retailers, real current prices, and it flagged the discount on the Corral pair without being asked. Satisfactory.
Six tests, six passes. Fable 5 handled vision, audio, code, creativity, and live research without any of the babysitting older models needed. The benchmark table says it is a step beyond Opus 4.8, and the everyday tests back that up. Though I am not sure when they will release the Fable model again for general use.