From 29fb1fdaa38078ee3be60bf36daabfc75d1a48a0 Mon Sep 17 00:00:00 2001
From: rakshith1928 <rakshithn6485@gmail.com>
Date: Sat, 15 Aug 2026 04:28:28 +0530
Subject: [PATCH] feat(perplexity-agent): add grok 4.6 and
 deepseek-v4-flash-0731 models configuration (#4736)

* feat(perplexity-agent): add grok 4.6 model configuration

* feat(perplexity-agent): add deepseek v4 flash model configuration
---
 .../deepseek/deepseek-v4-flash-0731.toml      | 14 ++++++++++++++
 .../perplexity-agent/models/xai/grok-4.6.toml | 19 +++++++++++++++++++
 2 files changed, 33 insertions(+)
 create mode 100644 providers/perplexity-agent/models/deepseek/deepseek-v4-flash-0731.toml
 create mode 100644 providers/perplexity-agent/models/xai/grok-4.6.toml

diff --git a/providers/perplexity-agent/models/deepseek/deepseek-v4-flash-0731.toml b/providers/perplexity-agent/models/deepseek/deepseek-v4-flash-0731.toml
new file mode 100644
index 000000000..d3f44c691
--- /dev/null
+++ b/providers/perplexity-agent/models/deepseek/deepseek-v4-flash-0731.toml
@@ -0,0 +1,14 @@
+# Perplexity Agent (OpenAI-compatible relay) pricing per:
+# https://docs.perplexity.ai/docs/agent-api/models#deepseek
+# Reasoning tokens billed at the output rate (no separate CoT price).
+# Effort: reasoning_effort = low|high|max
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.13
+output = 0.26
+cache_read = 0.028
diff --git a/providers/perplexity-agent/models/xai/grok-4.6.toml b/providers/perplexity-agent/models/xai/grok-4.6.toml
new file mode 100644
index 000000000..18e6bdcdd
--- /dev/null
+++ b/providers/perplexity-agent/models/xai/grok-4.6.toml
@@ -0,0 +1,19 @@
+# Pricing per Perplexity Agent API docs:
+# https://docs.perplexity.ai/docs/agent-api/models#xai
+# Context-based tiers: <=200k / >200k.
+base_model = "xai/grok-4.6"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.50
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4.00
+output = 12.00
+cache_read = 1.00