Gemini Pro News

Before 2025-12-01

Before 2025-12-15

Before 2025-11-21

Before 2026-01-01

Before 2025-11-14

GPT-5.3 (doesn't include GPT-5.3 Codex)

GPT-5.4 (includes GPT-5.4 Codex or other non-Pro variants)

Gemini 3.1 Flash Lite Preview

GPT-5.4 Pro

Grok 4.20 (out of beta)

Minimax M2.6/M2.7

Gemma 4

Haiku 4.6

Kimi K3

DeepSeek V4-Lite

Gemini 3.1 Flash Preview

GLM-5-Code

Gemini 3.1 Pro GA

Deepseek V4 (Full)

Opus 4.7

Gemini 3.1 Flash GA

Sonnet 5

Sonnet 4.7

Gemini 3.1 Flash Lite GA

Grok 5

Minimax M3

DeepSeek V3.3 or V3.5

GPT-5.3-Thinking

GPT-5.3-Pro

$10 - $12.5

$12.5 - $15

$15 - $17.5

<$2.5

$2.5 - $5

$5 - $7.5

$7.5 - $10

$17.5 - $20

$20 - $22.5

uses some non-AdamW optimizer

Some image input (multimodality)

>= 50% on TerminalBench (https://www.tbench.ai/leaderboard)

>= 60% on BrowseComp (https://www.kaggle.com/benchmarks/openai/browsecomp)

1M+ Context

>=25T pretraining tokens

>= 73% on SWE-Bench Verified (according to epoch.ai)

>= 2 shared experts

Some variation of NSA (Native Sparse Attention)

>=16 active experts

Gemini 2.5 Pro tier or higher on FictionBench (90.6%+ at 192k)

>=512 experts

intra-expert communication

DS-MoE with adaptative expert count

>= 44% on Humanity's Last Exam (text only) at scale.com leaderboard

DeepSeek reports some results with a full-blown deep research agent, and emphasizes that this is the intended use-mode

>=52B active parameters

>=1.5T parameters

YES

Other

6h30 - 7h00

6h00 - 6h30

5h30 - 6h00

4h30 - 5h00

5h00 - 5h30

7h00 - 7h30

3h30 - 4h00

4h00 - 4h30

NO

YES

< 0.05x

0.05x - 0.2x

0.2x - 0.5x

0.5x - 0.9x

0.9x - 1.1x

YES

YES

NO