Changelog

What changed in the data, and when. Entries are generated from the ingests themselves — a new model appearing at a provider, a published price moving, a benchmark or arena refresh landing — so this is a record of what we observed, not a list of announcements.

46 of 46 entries

July 2026

Price change · 25 Jul 2026

Qwen: Qwen3 30B A3B Instruct 2507 price down 44%

Blended 3:1 price moved from $0.15 to $0.08 per 1M tokens (OpenRouter).

Arena update · 25 Jul 2026

Arena Elo refreshed: 1002 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=376 text_sc=376 video=42 video_i2v=42

Arena update · 25 Jul 2026

Arena Elo refreshed: 1002 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=376 text_sc=376 video=42 video_i2v=42

Benchmark update · 25 Jul 2026

Benchmark refresh: 875 scores across 13 evaluations

Epoch AI benchmark archive (CC-BY 4.0), including LiveBench results redistributed by Epoch AI. The Known Good aggregates published results; it runs no evaluations.

Price change · 25 Jul 2026

MoonshotAI: Kimi K2.6 price down 15%

Blended 3:1 price moved from $1.37 to $1.16 per 1M tokens (OpenRouter).

Arena update · 25 Jul 2026

Arena Elo refreshed: 1002 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=376 text_sc=376 video=42 video_i2v=42

Benchmark update · 25 Jul 2026

Benchmark refresh: 875 scores across 13 evaluations

Epoch AI benchmark archive (CC-BY 4.0), including LiveBench results redistributed by Epoch AI. The Known Good aggregates published results; it runs no evaluations.

Model added · 25 Jul 2026

New model seen in arena leaderboards: deepseek-r1-0528

Model added · 25 Jul 2026

New model seen in arena leaderboards: deepseek-r1

Arena update · 25 Jul 2026

Arena Elo refreshed: 1002 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=376 text_sc=376 video=42 video_i2v=42

Benchmark update · 25 Jul 2026

Benchmark refresh: 875 scores across 13 evaluations

Epoch AI benchmark archive (CC-BY 4.0), including LiveBench results redistributed by Epoch AI. The Known Good aggregates published results; it runs no evaluations.

Model added · 25 Jul 2026

New model seen in arena leaderboards: chatgpt-image-latest-high-fidelity (20251216)

Model added · 25 Jul 2026

New model seen in arena leaderboards: gpt-image-2 (medium)

Model added · 25 Jul 2026

New model seen in arena leaderboards: gemini-3.1-flash-lite-image (nano-banana-2-lite)

Model added · 25 Jul 2026

New model seen in arena leaderboards: gemini-3.1-flash-image (nano-banana-2) [web-search]

Model added · 25 Jul 2026

New model seen in arena leaderboards: gemini-3-pro-image-preview (nano-banana-pro)

Model added · 25 Jul 2026

New model seen in arena leaderboards: gemini-3-pro-image-2k (nano-banana-pro)

Model added · 25 Jul 2026

New model seen in arena leaderboards: gemini-2.5-flash-image-preview (nano-banana)

Model added · 25 Jul 2026

New model seen in arena leaderboards: mimo-v2-flash (thinking)

Model added · 25 Jul 2026

New model seen in arena leaderboards: mimo-v2-flash (non-thinking)

Model added · 25 Jul 2026

New model seen in arena leaderboards: glm-5.2 (max)

Arena update · 25 Jul 2026

Arena Elo refreshed: 1004 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=377 text_sc=377 video=42 video_i2v=42

Model added · 25 Jul 2026

New model seen in arena leaderboards: claude-opus-4-20250514-thinking-16k

Model added · 25 Jul 2026

New model seen in arena leaderboards: claude-opus-4-1-20250805-thinking-16k

Model added · 25 Jul 2026

New model seen in arena leaderboards: claude-fable-5

Model added · 25 Jul 2026

New model seen in arena leaderboards: claude-3-7-sonnet-20250219-thinking-32k

Model added · 25 Jul 2026

New model seen in arena leaderboards: chatgpt-4o-latest-20250326

Model added · 25 Jul 2026

New model seen in arena leaderboards: chatglm3-6b

Model added · 25 Jul 2026

New model seen in arena leaderboards: chatglm2-6b

Model added · 25 Jul 2026

New model seen in arena leaderboards: chatglm-6b

Model added · 25 Jul 2026

New model seen in arena leaderboards: c4ai-aya-expanse-8b

Model added · 25 Jul 2026

New model seen in arena leaderboards: c4ai-aya-expanse-32b

Model added · 25 Jul 2026

New model seen in arena leaderboards: athene-v2-chat

Model added · 25 Jul 2026

New model seen in arena leaderboards: athene-70b-0725

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-26-02-10

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-26-01-10

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-12-10

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-11-10

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-10-20

Model added · 25 Jul 2026

New model seen in arena leaderboards: amazon-nova-experimental-chat-10-09

Model added · 25 Jul 2026

New model seen in arena leaderboards: alpaca-13b

Model added · 25 Jul 2026

New model seen in arena leaderboards: RWKV-4-Raven-14B

Arena update · 25 Jul 2026

Arena Elo refreshed: 1004 ratings across 7 arenas

image=73 image_edit=51 speech=42 text=377 text_sc=377 video=42 video_i2v=42

Model added · 25 Jul 2026

143 models added from Epoch AI benchmark data

See ingest_run note for the full list.

Benchmark update · 25 Jul 2026

Benchmark refresh: 874 scores across 13 evaluations

Epoch AI benchmark archive (CC-BY 4.0), including LiveBench results redistributed by Epoch AI. The Known Good aggregates published results; it runs no evaluations.

Model added · 25 Jul 2026

Now tracking 345 models

Initial catalogue import from OpenRouter.

What lands here. New models are recorded when a source first offers one — OpenRouter usually sees them first, so a model can appear here with sparse metadata and fill in later. Price moves are recorded against published rates. Benchmark and arena entries record that a publisher's refresh was ingested, never that we ran anything. A run that failed its sanity check is not shown as a change, because it never touched the live data — those are on the freshness table instead.