<!-- LLM_VERSION_INFO
FORMAT: text/markdown
CONTENT_TYPE: article
ORIGINAL_URL: https://www.vals.ai/models/fireworks_deepseek-v3p2-thinking
ALTERNATE_VERSION: models/fireworks_deepseek-v3p2-thinking/index.html (text/html)
EXTRACTION_DATE: 2026-04-17T00:42:53.368Z

This is the markdown version with text-only content (images converted to alt-text).
For rich formatting with images, request the HTML version at: models/fireworks_deepseek-v3p2-thinking/index.html
-->

# Open Weights & Proprietary Models

## Release Dates and Models

### 2026 Models

- **Claude Opus 4.7**  
    
  Release date: 4/16/2026

- **Muse Spark**  
    
  Release date: 4/8/2026

- **Gemma 4 31B IT**  
    
  Release date: 4/2/2026

- **Qwen 3.6 Plus**  
    
  Release date: 4/2/2026

- **GLM 5.1**  
    
  Release date: 4/1/2026

- **Trinity Large Thinking**  
    
  Release date: 3/17/2026

- **GPT 5.4 Mini**  
    
  Release date: 3/17/2026

- **GPT 5.4 Nano**  
    
  Release date: 3/17/2026

- **MiniMax-M2.7**  
    
  Release date: 3/9/2026

- **Grok 4.20 (Reasoning)**  
    
  Release date: 3/5/2026

- **GPT 5.4**  
    
  Release date: 3/3/2026

- **Gemini 3.1 Flash Lite Preview**  
    
  Release date: 2/24/2026

- **GPT 5.3 Codex**  
    
  Release date: 2/23/2026

- **Qwen 3.5 Flash**  
    
  Release date: 2/19/2026

- **Gemini 3.1 Pro Preview (02/26)**  
    
  Release date: 2/17/2026

- **Claude Sonnet 4.6**  
    
  Release date: 2/16/2026

- **Qwen 3.5 Plus**  
    
  Release date: 2/12/2026

- **MiniMax-M2.5**  
    
  Release date: 2/12/2026

- **MiniMax-M2.5**  
    
  Release date: 2/11/2026

- **GLM 5**  
    
  Release date: 2/5/2026

- **Claude Opus 4.6 (Nonthinking)**  
    
  Release date: 2/5/2026

- **Claude Opus 4.6 (Thinking)**  
    
  Release date: 1/26/2026

- **Kimi K2.5**  
    
  Release date: 1/23/2026

- **Qwen 3 Max Thinking**  
    
  Release date: 12/23/2025

- **MiniMax-M2.1**  
    
  Release date: 12/22/2025

- **GLM 4.7**  
    
  Release date: 12/17/2025

- **Gemini 3 Flash (12/25)**  
    
  Release date: 12/17/2025

- **MiMo V2 Flash**  
    
  Release date: 12/11/2025

- **GPT 5.2**  
    
  Release date: 12/11/2025

- **GPT 5.2 Codex**  
    
  Release date: 12/11/2025

## Model Details - DeepSeek V3.2 (Thinking)

- **Release Date:** 12/1/2025  
- **Default Provider:** Fireworks AI  
- **Accuracy (Vals Index):** 37.58% ± 1.50
- **Latency (Vals Index):** 831.80s
- **Cost/Test (Vals Index):** $0.24  
- **Context Window:** 160k  
- **Max Output Tokens:** 20k  
- **Input Modality:** Hyperparameter settings

### Hyperparameter Settings

- **Temperature:** 1  
- **Top P:** Default  
- **Top K:** Default  
- **Max Output Tokens:** 20,480  
- **Reasoning Effort:** high

### Benchmarks

#### Accuracy Rankings

- **Vals Index**  
---
- **CaseLaw (v2)**  
---
- **CorpFin**  
---
- **Finance Agent (v1.1)**  
---
- **ProofBench**  
---
- **TaxEval (v2)**  
---
- **Vibe Code Bench**  
---
- **AIME**  
---
- **GPQA**  
---
- **IOI**  
---
- **LiveCodeBench**  
---
- **LegalBench**  
---
- **MedQA**  
---
- **MMLU Pro**  
---
- **SWE-bench**  
---

---

### Contact Information  
Email: contact@vals.ai  
For access to proprietary benchmarks, contact us.